DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Continued Examination Under 37 CFR 1.114
A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on 06/30/2025 has been entered.
Information Disclosure Statement
The examiner has considered the information disclosure statements (IDS) submitted on 08/20/2025.
Response to Amendment
The amendment filed 06/30/2025 has been entered. Claims 1-2, 6-9, 11-12 and 16-19 remain pending in the application. Applicant's amendments to the claims have overcome the objections previously set forth in the Final Office Action mailed 03/31/2025.
Response to Arguments
Applicant’s arguments, filed 06/30/2025, with respect to the rejections of claims 1 and 11 under 103 have been fully considered and are persuasive because of the amendments. Therefore, the rejections have been withdrawn. However, upon further consideration, a new ground(s) of rejection is made in view of Lin et al. (NPL: KNOWLEDGE-GUIDED RECURRENT NEURAL NETWORK LEARNING FOR TASK-ORIENTED ACTION PREDICTION) and further in view of Tremblay et al. (US Pub. 2019/0228495).
Applicant’s arguments, filed 06/30/2025, with respect to the rejections of the claims under 101 have been fully considered and are not persuasive.
Applicant argues (page 6)
In the present response, claim 1 is further amended to recite "transmitting the task plan to the autonomous robot or vehicle." Applicants respectfully submit that the step of transmitting the task plan to the autonomous robot or vehicle is not an abstract process, and clarifies that amended claim 1, including the step of transmitting as well as the other technical details recited therein, is directed to a practical implementation and not merely an abstract idea.
Claim 11 has been amended in a similar fashion as claim 1, and is directed to patentable subject matter for at least the same reasons as claim 1.
In response
Claim 1 recites a method of generating a task plan for a task comprising the steps of “generating a search tree … generating the task plan … generating sequence data … acquiring context present in the sequence data and determining output sequence data”. These steps/limitations are based on observations, evaluations, judgments or opinion that are performable in the human mind or with the aid of pencil and paper, for example, a user can draw a tree with nodes representing the states relating to the task and edges representing actions to be performed to move from one state to the other, the user can generate a sequence of actions for achieving the task based on a suggested path from one node to a target node of the tree, the user can arrange the steps/actions to be performed in order, and the user can determine a certain feature/content in the data. Therefore, claim 1 recites mental processes.
Since the claim recites an abstract idea, additional limitations are analyzed under Step 2A Prong2 and 2B to determine if the additional limitations integrate the judicial exception into a practical application. According to MPEP 2106.05(a), limitations that are indicative of integration into a practical application when recited in a claim with a judicial exception include improvements to the functioning of a computer, or to any other technology or technical field.
The amended limitation “transmitting the task plan to the autonomous robot or vehicle” amounts to insignificant extra-solution activities of data transmitting, which does not amount to significantly more than the abstract idea (MPEP 2106.05(g)). The courts have found limitations directed to transmitting information electronically, recited at a high level of generality, to be well-understood, routine, and conventional (see MPEP 2106.05(d)(II), “receiving or transmitting data over a network”). Accordingly, this additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea.
The additional elements of “estimating a recommended path for an internal connection of the search tree by inputting the plurality of task states and the plurality of task actions to a neural network based on the search tree”, “performing, by the autonomous robot or vehicle, each of a plurality of tasks of the task plan associated with the recommended path to achieve the target state of the task”, “generating a trained neural network by training the neural network based on the plurality of task states and the plurality of task actions”, “estimating the recommended path based on the trained neural network” and “training the neural network based on the output sequence data”. These limitations are recited at a high-level of generality (i.e., as a generic device performing the generic computer functions) such that they amount no more than mere instructions to apply the exception using the generic computer components (MPEP 2106.05(f)). Accordingly, these additional elements do not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea.
The additional elements of “wherein the estimating of the recommended path comprises: …”, “wherein the generating of the trained neural network comprises: …”, “generating sequence data based on …” and “determining output sequence data based on …” amount to no more than generally linking the use of a judicial exception to a particular technological environment or field of use. As explained by the Supreme Court, a claim directed to a judicial exception cannot be made eligible "simply by having the applicant acquiesce to limiting the reach of the patent for the formula to a particular technological use." Diamond v. Diehr, 450 U.S. 175, 192 n.14, 209 USPQ 1, 10 n. 14 (1981). Thus, limitation that amounts to merely indicating a field of use or technological environment in which to apply a judicial exception does not integrate into a practical application (see MPEP 2106.05(h)).
Therefore, even when considering the elements in combination, the claim as a whole does not integrate the recited exception into a practical application.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-2, 6-9, 11-12 and 16-19 are rejected under 35 U.S.C. 101 because the claimed invention is directed to a judicial exception (i.e., a law of nature, a natural phenomenon, or an abstract idea) without significantly more.
Claim 1 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1:
The claim recites a method which falls within at least one of the four statutory categories of patent eligible subject matter.
Step 2:
Step 2A (prong 1):
The limitation of “generating a search tree based on a plurality of task states of the task and a plurality of task actions capable of being performed by the autonomous robot or vehicle for performing the task”, as drafted, is a process that, under its broadest reasonable interpretation, covers performance of the limitation in the mind. For example, “generating” in the context of this claim encompasses the user draws a tree with nodes representing the states relating to the task and edges representing actions to be performed to move from one state to the other.
The limitation of “generating the task plan by determining a target path that reaches from an initial state of the task to a target state of the task based on the recommended path”, as drafted, is a process that, under its broadest reasonable interpretation, covers performance of the limitation in the mind. For example, “generating” in the context of this claim encompasses the user generates a sequence of actions for achieving the task based on a suggested path from one node to a target node.
The limitation of “generating sequence data …”, as drafted, is a process that, under its broadest reasonable interpretation, covers performance of the limitation in the mind. For example, “generating” in the context of this claim encompasses the user determines the steps/actions to be performed in order.
The limitation of “acquiring context present in the sequence data”, as drafted, is a process that, under its broadest reasonable interpretation, covers performance of the limitation in the mind. For example, “acquiring” in the context of this claim encompasses the user determines a certain feature/content in the generated data.
The limitation of “determining output sequence data”, as drafted, is a process that, under its broadest reasonable interpretation, covers performance of the limitation in the mind. For example, “determining” in the context of this claim encompasses the user determines the data that arranged in order.
Therefore, the claim recites mental processes.
Step 2A (prong 2):
This judicial exception is not integrated into a practical application. In particular, the claim recites the additional elements of “an autonomous robot or vehicle”, “a neural network” and “a trained neural network”. The additional element is recited at a high-level of generality (i.e., as a generic device performing the generic computer functions) such that they amount no more than mere instructions to apply the exception using the generic computer components (MPEP 2106.05(f)). Accordingly, these additional elements do not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea.
The claim recites the additional elements of “estimating a recommended path for an internal connection of the search tree by inputting the plurality of task states and the plurality of task actions to a neural network based on the search tree”, “performing, by the autonomous robot or vehicle, each of a plurality of tasks of the task plan associated with the recommended path to achieve the target state of the task”, “generating a trained neural network by training the neural network based on the plurality of task states and the plurality of task actions”, “estimating the recommended path based on the trained neural network” and “training the neural network based on the output sequence data”. These limitations are recited at a high-level of generality (i.e., as a generic device performing the generic computer functions) such that they amount no more than mere instructions to apply the exception using the generic computer components (MPEP 2106.05(f)). Accordingly, these additional element do not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea.
The additional element of “transmitting the task plan to the autonomous robot or vehicle” amounts to insignificant extra-solution activities of data transmitting, which does not amount to significantly more than the abstract idea (MPEP 2106.05(g)). Accordingly, this additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea.
The additional elements of “wherein the estimating of the recommended path comprises:”, “wherein the generating of the trained neural network comprises:”, “generating sequence data based on …” and “determining output sequence data based on …” amount to no more than generally linking the use of a judicial exception to a particular technological environment or field of use. As explained by the Supreme Court, a claim directed to a judicial exception cannot be made eligible "simply by having the applicant acquiesce to limiting the reach of the patent for the formula to a particular technological use." Diamond v. Diehr, 450 U.S. 175, 192 n.14, 209 USPQ 1, 10 n. 14 (1981). Thus, limitation that amounts to merely indicating a field of use or technological environment in which to apply a judicial exception does not integrate into a practical application (see MPEP 2106.05(h)).
Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
As discussed above with respect to integration of the abstract idea into a practical application, the additional elements of “estimating a recommended path for an internal connection of the search tree by inputting the plurality of task states and the plurality of task actions to a neural network based on the search tree”, “performing, by the autonomous robot or vehicle, each of a plurality of tasks of the task plan associated with the recommended path to achieve the target state of the task”, “generating a trained neural network by training the neural network based on the plurality of task states and the plurality of task actions”, “estimating the recommended path based on the trained neural network” and “training the neural network based on the output sequence data” mount to no more than mere instructions to apply the exception using a generic computer component. Mere instructions to apply an exception using a generic computer component cannot provide an inventive concept.
The additional element of “transmitting the task plan to the autonomous robot or vehicle” is recited at a high level of generality and amounts to insignificant extra-solution activity related to mere data transmitting (MPEP 2106.05(g)). The courts have found limitations directed to transmitting information electronically, recited at a high level of generality, to be well-understood, routine, and conventional (see MPEP 2106.05(d)(II), “receiving or transmitting data over a network”).
The additional elements of “wherein the estimating of the recommended path comprises:”, “wherein the generating of the trained neural network comprises:”, “generating sequence data based on …” and “determining output sequence data based on …” amount to no more than generally linking the use of a judicial exception to a particular technological environment or field of use. As explained by the Supreme Court, a claim directed to a judicial exception cannot be made eligible "simply by having the applicant acquiesce to limiting the reach of the patent for the formula to a particular technological use." Diamond v. Diehr, 450 U.S. 175, 192 n.14, 209 USPQ 1, 10 n. 14 (1981). Thus, limitations that amount to merely indicating a field of use or technological environment in which to apply a judicial exception does not amount to significantly more than the judicial exception (see MPEP 2106.05(h)).
Therefore, the claim is not patent eligible under 35 U.S.C. 101.
Claim 2 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1:
The claim recites a method which falls within at least one of the four statutory categories of patent eligible subject matter.
Step 2:
Step 2A (prong 1):
The limitation of “generating nodes corresponding to the plurality of task states”, as drafted, is a process that, under its broadest reasonable interpretation, covers performance of the limitation in the mind. For example, “generating” in the context of this claim encompasses the user draws multiple nodes on the search tree, each node representing a task state.
The limitation of “generating the search tree by connecting the nodes via edges corresponding to the plurality of task actions”, as drafted, is a process that, under its broadest reasonable interpretation, covers performance of the limitation in the mind. For example, “generating” in the context of this claim encompasses the user draws a tree with multiple nodes, and edges connecting the nodes that representing actions to be performed to move from one node to the other.
Therefore, the claim recites mental processes.
Step 2A (prong 2):
The claim does not recite additional elements that integrate the judicial exception into a practical application.
Step 2B:
The claim does not recite additional elements that amount to significantly more than the judicial exception.
Claim 6 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1:
The claim recites a method which falls within at least one of the four statutory categories of patent eligible subject matter.
Step 2:
Step 2A (prong 1):
The limitation of “acquiring a hash code by performing a hash operation on a task state of the sequence data” recites a mathematical concept.
The limitation of “generating an information vector by encoding a task action and a task of the sequence data”, as drafted, is a process that, under its broadest reasonable interpretation, covers performance of the limitation in the mind. For example, “generating” in the context of this claim encompasses the user converts/translates data and generates a vector with the converted data.
The limitation of “generating the training data …”, as drafted, is a process that, under its broadest reasonable interpretation, covers performance of the limitation in the mind. For example, “generating” in the context of this claim encompasses the user creates sample data for further analysis.
Step 2:
Step 2A (prong 2):
This judicial exception is not integrated into a practical application. The additional element of “generating the training data based on the hash code and the information vector” amounts to no more than generally linking the use of a judicial exception to a particular technological environment or field of use. As explained by the Supreme Court, a claim directed to a judicial exception cannot be made eligible "simply by having the applicant acquiesce to limiting the reach of the patent for the formula to a particular technological use." Diamond v. Diehr, 450 U.S. 175, 192 n.14, 209 USPQ 1, 10 n. 14 (1981). Thus, limitation that amounts to merely indicating a field of use or technological environment in which to apply a judicial exception does not integrate into a practical application (see MPEP 2106.05(h)).
Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
As discussed above with respect to integration of the abstract idea into a practical application, the additional element of “generating the training data based on the hash code and the information vector” amounts to no more than generally linking the use of a judicial exception to a particular technological environment or field of use. As explained by the Supreme Court, a claim directed to a judicial exception cannot be made eligible "simply by having the applicant acquiesce to limiting the reach of the patent for the formula to a particular technological use." Diamond v. Diehr, 450 U.S. 175, 192 n.14, 209 USPQ 1, 10 n. 14 (1981). Thus, limitation that amounts to merely indicating a field of use or technological environment in which to apply a judicial exception does not amount to significantly more than the judicial exception (see MPEP 2106.05(h)).
Therefore, the claim is not patent eligible under 35 U.S.C. 101.
Claim 7 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1:
The claim recites a method which falls within at least one of the four statutory categories of patent eligible subject matter.
Step 2:
Step 2A (prong 1):
The limitation of “performing one-hot encoding on the task action and the task”, as drafted, is a process that, under its broadest reasonable interpretation, covers performance of the limitation in the mind. For example, “performing” in the context of this claim encompasses the user presents/describes the categorical variables as binary format.
Step 2A (prong 2):
This judicial exception is not integrated into a practical application. The additional element of “wherein the generating of the information vector comprises acquiring a one-hot vector as the information vector by performing …” amounts to no more than generally linking the use of a judicial exception to a particular technological environment or field of use. As explained by the Supreme Court, a claim directed to a judicial exception cannot be made eligible "simply by having the applicant acquiesce to limiting the reach of the patent for the formula to a particular technological use." Diamond v. Diehr, 450 U.S. 175, 192 n.14, 209 USPQ 1, 10 n. 14 (1981). Thus, limitation that amounts to merely indicating a field of use or technological environment in which to apply a judicial exception does not integrate into a practical application (see MPEP 2106.05(h)).
Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
As discussed above with respect to integration of the abstract idea into a practical application, the additional element of “wherein the generating of the information vector comprises acquiring a one-hot vector as the information vector by performing …” amounts to no more than generally linking the use of a judicial exception to a particular technological environment or field of use. As explained by the Supreme Court, a claim directed to a judicial exception cannot be made eligible "simply by having the applicant acquiesce to limiting the reach of the patent for the formula to a particular technological use." Diamond v. Diehr, 450 U.S. 175, 192 n.14, 209 USPQ 1, 10 n. 14 (1981). Thus, limitation that amounts to merely indicating a field of use or technological environment in which to apply a judicial exception does not amount to significantly more than the judicial exception (see MPEP 2106.05(h)).
Therefore, the claim is not patent eligible under 35 U.S.C. 101.
Claim 8 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1:
The claim recites a method which falls within at least one of the four statutory categories of patent eligible subject matter.
Step 2:
Step 2A (prong 1):
The limitation of “determining an edge connected to a front node in the search tree based on the recommended path”, as drafted, is a process that, under its broadest reasonable interpretation, covers performance of the limitation in the mind. For example, “determining” in the context of this claim encompasses the user determines a first node of the suggested path, and an edge connecting the first node to other node(s).
The limitation of “determining a child node connected to the edge based on the edge”, as drafted, is a process that, under its broadest reasonable interpretation, covers performance of the limitation in the mind. For example, “determining” in the context of this claim encompasses the user, based on the suggested path, determines the node(s) that connect to the first node.
Therefore, the claim recites mental processes.
Step 2A (prong 2):
The claim does not recite additional elements that integrate the judicial exception into a practical application.
Step 2B:
The claim does not recite additional elements that amount to significantly more than the judicial exception.
Claim 9 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1:
The claim recites a method which falls within at least one of the four statutory categories of patent eligible subject matter.
Step 2:
Step 2A (prong 1):
The limitation of “determining a recommended action type among the plurality of task actions based on the recommended path”, as drafted, is a process that, under its broadest reasonable interpretation, covers performance of the limitation in the mind. For example, “determining” in the context of this claim encompasses the user, based on a suggested path, determines which actions will be performed to complete the suggested path.
The limitation of “determining the edge based on the recommended action type”, as drafted, is a process that, under its broadest reasonable interpretation, covers performance of the limitation in the mind. For example, “determining” in the context of this claim encompasses the user, based on the recommended actions, determines the edges that connects the nodes associated with those actions.
Therefore, the claim recites mental processes.
Step 2A (prong 2):
The claim does not recite additional elements that integrate the judicial exception into a practical application.
Step 2B:
The claim does not recite additional elements that amount to significantly more than the judicial exception.
Claim 11 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1:
The claim recites a system which falls within at least one of the four statutory categories of patent eligible subject matter.
Step 2:
Step 2A (prong 1):
The limitation of “generate a search tree based on a plurality of task states of the task and a plurality of task actions for performing the task”, as drafted, is a process that, under its broadest reasonable interpretation, covers performance of the limitation in the mind. For example, “generate” in the context of this claim encompasses the user draws a tree with nodes representing the states relating to the task and edges representing actions to be performed to move from one state to the other.
The limitation of “generate the task plan by determining a target path that reaches from an initial state of the task to a target state of the task based on the recommended path”, as drafted, is a process that, under its broadest reasonable interpretation, covers performance of the limitation in the mind. For example, “generate” in the context of this claim encompasses the user generates a sequence of actions for achieving the task based on a suggested path from one node to a target node.
The limitation of “generate sequence data …”, as drafted, is a process that, under its broadest reasonable interpretation, covers performance of the limitation in the mind. For example, “generate” in the context of this claim encompasses the user determines the steps/actions to be performed in order.
The limitation of “acquire context present in the sequence data”, as drafted, is a process that, under its broadest reasonable interpretation, covers performance of the limitation in the mind. For example, “acquire” in the context of this claim encompasses the user determines a certain feature in the generated data.
The limitation of “determine output sequence data”, as drafted, is a process that, under its broadest reasonable interpretation, covers performance of the limitation in the mind. For example, “determine” in the context of this claim encompasses the user determines the data that arranged in order.
Therefore, the claim recites mental processes.
Step 2A (prong 2):
This judicial exception is not integrated into a practical application. In particular, the claim recites the additional elements of “an autonomous robot or vehicle”, “a processor”, “a memory”, “a neural network” and “a trained neural network”. The additional element is recited at a high-level of generality (i.e., as a generic device performing the generic computer functions) such that they amount no more than mere instructions to apply the exception using the generic computer components (MPEP 2106.05(f)). Accordingly, these additional elements do not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea.
The claim recites the additional elements of “estimate a recommended path for an internal connection of the search tree by inputting the plurality of task states and the plurality of task actions to a neural network based on the search tree”, “perform, by the autonomous robot or vehicle, each of a plurality of tasks of the task plan associated with the recommended path to achieve the target state of the task”, “generate a trained neural network by training the neural network based on the plurality of task states and the plurality of task actions, and estimate the recommended path based on the trained neural network” and “train the neural network based on the output sequence data”. These limitations are recited at a high-level of generality (i.e., as a generic device performing the generic computer functions) such that they amount no more than mere instructions to apply the exception using the generic computer components (MPEP 2106.05(f)). Accordingly, this additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea.
The additional element of “transmit the task plan to the autonomous robot or vehicle” amounts to insignificant extra-solution activities of data transmitting, which does not amount to significantly more than the abstract idea (MPEP 2106.05(g)). Accordingly, this additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea.
The additional elements of “generate sequence data based on …” and “determine output sequence data based on …” amount to no more than generally linking the use of a judicial exception to a particular technological environment or field of use. As explained by the Supreme Court, a claim directed to a judicial exception cannot be made eligible "simply by having the applicant acquiesce to limiting the reach of the patent for the formula to a particular technological use." Diamond v. Diehr, 450 U.S. 175, 192 n.14, 209 USPQ 1, 10 n. 14 (1981). Thus, limitations that amount to merely indicating a field of use or technological environment in which to apply a judicial exception does not integrate into a practical application (see MPEP 2106.05(h)).
Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
As discussed above with respect to integration of the abstract idea into a practical application, the additional elements of “an autonomous robot or vehicle”, “a processor”, “a memory”, “a neural network” and “a trained neural network” to perform the generic computer functions amount to no more than mere instructions to apply the exception using a generic computer component. Mere instructions to apply an exception using a generic computer component cannot provide an inventive concept.
The additional elements of “estimate a recommended path for an internal connection of the search tree by inputting the plurality of task states and the plurality of task actions to a neural network based on the search tree”, “perform, by the autonomous robot or vehicle, each of a plurality of tasks of the task plan associated with the recommended path to achieve the target state of the task”, “generate a trained neural network by training the neural network based on the plurality of task states and the plurality of task actions, and estimate the recommended path based on the trained neural network” and “train the neural network based on the output sequence data” mount to no more than mere instructions to apply the exception using a generic computer component. Mere instructions to apply an exception using a generic computer component cannot provide an inventive concept.
The additional element of “transmit the task plan to the autonomous robot or vehicle” is recited at a high level of generality and amounts to insignificant extra-solution activity related to mere data transmitting (MPEP 2106.05(g)). The courts have found limitations directed to transmitting information electronically, recited at a high level of generality, to be well-understood, routine, and conventional (see MPEP 2106.05(d)(II), “receiving or transmitting data over a network”).
The additional elements of “generate sequence data based on …” and “determine output sequence data based on …” amount to no more than generally linking the use of a judicial exception to a particular technological environment or field of use. As explained by the Supreme Court, a claim directed to a judicial exception cannot be made eligible "simply by having the applicant acquiesce to limiting the reach of the patent for the formula to a particular technological use." Diamond v. Diehr, 450 U.S. 175, 192 n.14, 209 USPQ 1, 10 n. 14 (1981). Thus, limitations that amount to merely indicating a field of use or technological environment in which to apply a judicial exception does not amount to significantly more than the judicial exception (see MPEP 2106.05(h)).
Therefore, the claim is not patent eligible under 35 U.S.C. 101.
Claim 12 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1:
The claim recites a system which falls within at least one of the four statutory categories of patent eligible subject matter.
Step 2:
Step 2A (prong 1):
The limitation of “generate nodes corresponding to the plurality of task states”, as drafted, is a process that, under its broadest reasonable interpretation, covers performance of the limitation in the mind. For example, “generating” in the context of this claim encompasses the user draws multiple nodes on the search tree, each node representing a task state.
The limitation of “generate the search tree by connecting the nodes via edges corresponding to the plurality of task actions”, as drafted, is a process that, under its broadest reasonable interpretation, covers performance of the limitation in the mind. For example, “generating” in the context of this claim encompasses the user draws a tree with multiple nodes, and edges connecting the nodes that representing actions to be performed to move from one node to the other.
Therefore, the claim recites mental processes.
Step 2A (prong 2):
The claim does not recite additional elements that integrate the judicial exception into a practical application.
Step 2B:
The claim does not recite additional elements that amount to significantly more than the judicial exception.
Claim 16 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1:
The claim recites a system which falls within at least one of the four statutory categories of patent eligible subject matter.
Step 2:
Step 2A (prong 1):
The limitation of “acquire a hash code by performing a hash operation on a task state of the sequence data” recites a mathematical concept.
The limitation of “generate an information vector by encoding a task action and a task of the sequence data”, as drafted, is a process that, under its broadest reasonable interpretation, covers performance of the limitation in the mind. For example, “generate” in the context of this claim encompasses the user converts/translates data and generates a vector with the converted data.
The limitation of “generate the training data …”, as drafted, is a process that, under its broadest reasonable interpretation, covers performance of the limitation in the mind. For example, “generate” in the context of this claim encompasses the user creates sample data for further analysis.
Step 2:
Step 2A (prong 2):
This judicial exception is not integrated into a practical application. The additional element of “generate the training data based on the hash code and the information vector” amounts to no more than generally linking the use of a judicial exception to a particular technological environment or field of use. As explained by the Supreme Court, a claim directed to a judicial exception cannot be made eligible "simply by having the applicant acquiesce to limiting the reach of the patent for the formula to a particular technological use." Diamond v. Diehr, 450 U.S. 175, 192 n.14, 209 USPQ 1, 10 n. 14 (1981). Thus, limitation that amounts to merely indicating a field of use or technological environment in which to apply a judicial exception does not integrate into a practical application (see MPEP 2106.05(h)).
Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
As discussed above with respect to integration of the abstract idea into a practical application, the additional element of “generate the training data based on the hash code and the information vector” amounts to no more than generally linking the use of a judicial exception to a particular technological environment or field of use. As explained by the Supreme Court, a claim directed to a judicial exception cannot be made eligible "simply by having the applicant acquiesce to limiting the reach of the patent for the formula to a particular technological use." Diamond v. Diehr, 450 U.S. 175, 192 n.14, 209 USPQ 1, 10 n. 14 (1981). Thus, limitation that amounts to merely indicating a field of use or technological environment in which to apply a judicial exception does not amount to significantly more than the judicial exception (see MPEP 2106.05(h)).
Therefore, the claim is not patent eligible under 35 U.S.C. 101.
Claim 17 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1:
The claim recites a system which falls within at least one of the four statutory categories of patent eligible subject matter.
Step 2:
Step 2A (prong 1):
The limitation of “perform one-hot encoding on the task action and the task”, as drafted, is a process that, under its broadest reasonable interpretation, covers performance of the limitation in the mind. For example, “perform” in the context of this claim encompasses the user presents/describes the categorical variables as binary format.
Step 2A (prong 2):
This judicial exception is not integrated into a practical application. The additional element of “acquire a one-hot vector as the information vector by performing …” amounts to no more than generally linking the use of a judicial exception to a particular technological environment or field of use. As explained by the Supreme Court, a claim directed to a judicial exception cannot be made eligible "simply by having the applicant acquiesce to limiting the reach of the patent for the formula to a particular technological use." Diamond v. Diehr, 450 U.S. 175, 192 n.14, 209 USPQ 1, 10 n. 14 (1981). Thus, limitation that amounts to merely indicating a field of use or technological environment in which to apply a judicial exception does not integrate into a practical application (see MPEP 2106.05(h)).
Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
As discussed above with respect to integration of the abstract idea into a practical application, the additional element of “acquire a one-hot vector as the information vector by performing …” amounts to no more than generally linking the use of a judicial exception to a particular technological environment or field of use. As explained by the Supreme Court, a claim directed to a judicial exception cannot be made eligible "simply by having the applicant acquiesce to limiting the reach of the patent for the formula to a particular technological use." Diamond v. Diehr, 450 U.S. 175, 192 n.14, 209 USPQ 1, 10 n. 14 (1981). Thus, limitation that amounts to merely indicating a field of use or technological environment in which to apply a judicial exception does not amount to significantly more than the judicial exception (see MPEP 2106.05(h)).
Therefore, the claim is not patent eligible under 35 U.S.C. 101.
Claim 18 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1:
The claim recites a system which falls within at least one of the four statutory categories of patent eligible subject matter.
Step 2:
Step 2A (prong 1):
The limitation of “determine an edge connected to a front node in the search tree based on the recommended path”, as drafted, is a process that, under its broadest reasonable interpretation, covers performance of the limitation in the mind. For example, “determine” in the context of this claim encompasses the user determines a first node of the suggested path, and an edge connecting the first node to other node(s).
The limitation of “determine a child node connected to the edge based on the edge”, as drafted, is a process that, under its broadest reasonable interpretation, covers performance of the limitation in the mind. For example, “determine” in the context of this claim encompasses the user, based on the suggested path, determines the node(s) that connect to the first node.
Therefore, the claim recites mental processes.
Step 2A (prong 2):
The claim does not recite additional elements that integrate the judicial exception into a practical application.
Step 2B:
The claim does not recite additional elements that amount to significantly more than the judicial exception.
Claim 19 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1:
The claim recites a system which falls within at least one of the four statutory categories of patent eligible subject matter.
Step 2:
Step 2A (prong 1):
The limitation of “determine a recommended action type among the plurality of task actions based on the recommended path”, as drafted, is a process that, under its broadest reasonable interpretation, covers performance of the limitation in the mind. For example, “determine” in the context of this claim encompasses the user, based on a suggested path, determines which actions will be performed to complete the suggested path.
The limitation of “determine the edge based on the recommended action type”, as drafted, is a process that, under its broadest reasonable interpretation, covers performance of the limitation in the mind. For example, “determine” in the context of this claim encompasses the user, based on the recommended actions, determines the edges that connects the nodes associated with those actions.
Therefore, the claim recites mental processes.
Step 2A (prong 2):
The claim does not recite additional elements that integrate the judicial exception into a practical application.
Step 2B:
The claim does not recite additional elements that amount to significantly more than the judicial exception.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1-2, 8-9, 11-12 and 18-19 are rejected under 35 U.S.C. 103 as being unpatentable over Lin et al. (NPL: KNOWLEDGE-GUIDED RECURRENT NEURAL NETWORK LEARNING FOR TASK-ORIENTED ACTION PREDICTION) in view of Tremblay et al. (US Pub. 2019/0228495).
As per claim 1, Lin teaches a method of generating a task plan for a task [Fig. 1, “make tea” task] performed by an autonomous robot or vehicle [Fig. 1, page 1, introduction, “Automatically predicting and executing a sequence of actions given a specific task would be one quite expected ability for intelligent robots”; page 6, Col. 1, section 7, 1st paragraph, “In this paper, we address a challenging problem, i.e., predicting a sequence of actions to accomplish a specific task under a certain scene”; Examiner interprets “predicting a sequence of actions” to accomplish a specific task as predicting “predicting/generating a task plan” for a task], the method comprising:
generating a search tree based on a plurality of task states of the task and a plurality of task actions capable of being performed by the autonomous robot or vehicle for performing the task [Fig. 2, page 2, Col. 1, 2nd paragraph, “we present a two-stage training method by employing a temporal And-Or graph (AOG) representation [5, 6]. First, we define the AOG for task description, which hierarchically decomposes a task into atomic actions according to their temporal dependencies. In this semantic representation, an and-node represents the chronological decomposition of a task (or sub-task); an or-node represents the alternative ways to complete the certain task (or sub-task); leaf-nodes represent the pre-defined atomic actions. The AOG can thus contain all possible action sequences for each task”; Examiner interprets the chronological decomposition of a task or sub-task comprises a state, interprets the graph comprising the states connected by the atomic actions as a search tree based on the task states of the task and a plurality of task actions, and interprets decomposing a task into atomic actions to define the AOG as generating a search tree based on the task states of the task and a plurality of task actions, and in the introduction, Lin teaches the intelligent robots are capable of performing the actions];
estimating a recommended path for an internal connection of the search tree by inputting the plurality of task states and the plurality of task actions to a neural network based on the search tree [Fig. 4, page 4, Col. 1, 2nd and 3rd paragraphs, “Specifically, we collect a small set of samples annotated with the selections of all or-nodes given a scene image for each task and define the cross-entropy objective function to train the AOG-LSTM. Once the AOG-LSTM is trained, we adopt it to predict the sub-branch selections for all the or-nodes in the And-Or graph given different scene images, and generate the corresponding action sequences”; The examiner notes that, as written, the limitation is directed to path finding for a subset of the search tree or a subset of the arbitrary task. Page 8, lines 4-8 of the instant specification recites "The neural network (or an artificial neural network) may include a statistical training algorithm that mimics a biological nerve in machine training and cognitive science. The neural network may refer to a general model that has the ability to solve a problem, where artificial neurons (nodes) forming the network through synaptic combinations change a connection strength of synapses through training", page 8, line 13-page 9, line 2 of the instant specification lists example neural networks (including "CNN", "RNN" and "LSTM"), and page 17, lines 14-15 of the instant specification recites "The neural network may include an input layer 510, a hidden layer 530 and an output layer 550", but the instant specification does not appear to explicitly define a neural network. As such, the broadest reasonable interpretation of neural network based on the search tree, as written and in light of the specification, includes LIN's LSTM network trained on samples drawn from the AOG. Predicting a sub- branch is reasonably understood to be encompassed by estimating a recommended path for an internal connection of the search tree];
generating the task plan by determining a target path that reaches from an initial state of the task to a target state of the task based on the recommended path [Fig. 1 shows 2 different task plans each comprising a sequence of actions for completing the task “make tea”, where, an initial state S0 of the task may describe a first location of the robot, and the target state st is the state after taking an action “pour into, cup”; page 5, section 6.2.1, 3rd paragraph, "Then we evaluate the accuracy of generating the action sequences, i.e., whether the task is completed successfully. We define the sequence accuracy as the fraction of complete correct sequences with respect to all predicted sequences"; As discussed above, LIN teaches generating a task plan for a task utilizing a neural network which outputs predicted next actions, or recommended paths. Page 11, lines 9-11 of the instant specification recites "The processor 100 may generate the task plan by determining a target path that reaches from an initial state of the task to a target state of the task based on the recommended path", but does not appear to explicitly define a target path. As such, as written and in light of the specification, the broadest reasonable interpretation of generating the task plan by determining a target path that reaches from an initial state of the task to a target state of the task based on the recommended path includes assessing the accuracy of the recommended path in achieving the goal of the task];
performing, by the autonomous robot or vehicle, each of a plurality of tasks of the task plan associated with the recommended path to achieve the target state of the task [Fig. 1 shows 2 different task plans each comprising a sequence of actions (perform by a robot) for completing the task “make tea”. Introduction, 1st paragraph, “executing a sequence of actions given a specific task would be one quite expected ability for intelligent robots. For example, to complete the task of “make tea” under the scene shown in Figure 1, an agent needs to plan and successively execute a number of steps, e.g., “move to the tea box”, “grasp the tea box”. In this paper we aim to train a neural network model to enable such a capability; where, the agent includes "intelligent robots"]; and
transmitting the task plan to the autonomous robot or vehicle [Fig. 1 shows 2 different task plans each comprising a sequence of actions (perform by a robot) for completing the task “make tea”. Introduction, 1st paragraph, “executing a sequence of actions given a specific task would be one quite expected ability for intelligent robots. For example, to complete the task of “make tea” under the scene shown in Figure 1, an agent needs to plan and successively execute a number of steps, e.g., “move to the tea box”, “grasp the tea box”, where, the agent includes "intelligent robots"; It can be seen that the task plan is transmitted to the robot for executing],
wherein the estimating of the recommended path comprises: generating a trained neural network by training the neural network based on the plurality of task states and the plurality of task actions [page 1, introduction, 1st paragraph, “to complete the task of “make tea” under the scene shown in Figure 1, an agent needs to plan and successively execute a number of steps, e.g., “move to the tea box”, “grasp the tea box”. In this paper we aim to train a neural network model to enable such a capability”; page 2, Col. 1, 2nd paragraph, “we train an auxiliary LSTM network (named AOG-LSTM) to predict the selection at the or-nodes in the AOG, and can thus produce a large number of new valid samples (i.e., task-oriented action sequences) that can be used for training the Action-LSTM. Notably, training the AOG-LSTM requires only a few manually annotated samples (i.e., scene images and the corresponding action sequences)”; It can be seen that Scene images and corresponding action sequences are reasonably understood to include a plurality of task states and a plurality of task actions]; and
estimating the recommended path based on the trained neural network [page 6, section 7, 1st paragraph, “In this paper, we address a challenging problem, i.e., predicting a sequence of actions to accomplish a specific task under a certain scene, by developing a recurrent LSTM neural network”; It can be seen that predicting a sequence of actions to accomplish a specific task is reasonably understood to include estimating a recommended path], and
wherein the generating of the trained neural network comprises: generating sequence data based on the plurality of task states, the plurality of task actions, and the task [page 6, section 7, 1st paragraph, “In this paper, we address a challenging problem, i.e., predicting a sequence of actions to accomplish a specific task under a certain scene, by developing a recurrent LSTM neural network”; Fig. 1 shows 2 different task plans each comprising a sequence of actions (sequence data) for completing the task “make tea”; It can be seen that the 2 task plans are generated based on the states such as the first location of a robot before taking any action from a list of actions such as: “move to, grasp, open, etc., and the actions take to complete the “make tea” task];
Lin does not explicitly teach
acquiring context present in the sequence data,
determining output sequence data based on the context, and the sequence data, and
training the neural network based on the output sequence data.
Tremblay teaches
acquiring context present in the sequence data [abstract, “receive data involving the performance of a task in the physical world. The data can be provided as input to a perception network to infer a set of percepts about the task … The percepts can be provided as input to a plan generation network, which can infer a set of actions as part of a plan … The plan can be reviewed and any corrections made … Once the plan is verified as correct, the plan (and any related data) can be provided as input to an execution network that can infer instructions to cause the robot, and/or another robot, to perform the task”; paragraph 0026, “Once a plan is generated … The user can view the plan to determine whether the plan accurately represents the task to be performed”; Lin Fig. 1 shows 2 different task plans each comprising a sequence of actions (sequence data) for completing the task “make tea”, while Tremblay teaches once the task plan is generated, the plan is viewed to determine if it accurately represents the task to be performed. It can be seen that the process of viewing the plan to determine if it accurately represents the task to be performed, comprises viewing the context/content of the plan/actions result to determining if the task is accurately performed based on the plan/actions, therefore, the combination of Lin and Tremblay teaches the above limitation],
determining output sequence data based on the context, and the sequence data [paragraph 0026, “Once a plan is generated … The user can view the plan to determine whether the plan accurately represents the task to be performed. If not, the user can instruct the robot to capture data for another performance of the task using the objects 120, or can manually correct the plan to correctly represent the task to be performed, among other such options”; It can be seen that the generated plan which comprises task actions is viewed and corrected if needed to generate a corrected plan comprises task actions to accurately perform the task], and
training the neural network based on the output sequence data [paragraph 0026, “Once a plan is generated … The user can view the plan to determine whether the plan accurately represents the task to be performed … After the plan is verified, the plan and/or associated data can be processed using an execution neural network, for example, to generate a set of instructions executable by the robot 102 to perform the task”].
It would have been obvious to one of ordinary skill in the art before the effective filing date of the invention to have modified the method of predicting a sequence of actions to complete a task of Lin to include the processes of acquiring context present in the sequence data, determining output sequence data, and training the neural network based on the output sequence data of Tremblay. Doing so would help predicting, by the neural network, a set of instructions executable by the robot to perform the task (Tremblay, 0026).
As per claim 2, Lin and Tremblay teach the method of claim 1.
Lin further teaches
generating nodes corresponding to the plurality of task states [page 3, Fig. 2(a)
PNG
media_image1.png
220
372
media_image1.png
Greyscale
It can be seen that the “and” nodes, “or” nodes, and leaf nodes of LI N's parse tree are reasonably understood to be encompassed by nodes corresponding to the plurality of task states]
generating the search tree by connecting the nodes via edges corresponding to the plurality of task actions [Figs. 1-2, page 2, Col. 1, 2nd paragraph, “In this semantic representation, an and-node represents the chronological decomposition of a task (or sub-task); an or-node represents the alternative ways to complete the certain task (or sub-task); leaf-nodes represent the pre-defined atomic actions”; It can be seen that alternate ways to complete the certain task and pre-defined atomic actions are reasonably understood to include actions].
As per claim 8, Lin and Tremblay teach the method of claim 1.
Lin further teaches
the generating of the task plan comprises: determining an edge connected to a front node in the search tree based on the recommended path [Fig. 2, page 3, section 4.2, 2nd paragraph – page 4, Col. 1, 2nd paragraph, “which serves as the initial hidden state of the AOG-LSTM. It then encodes the initial AOG as an adjacency matrix. The matrix is further rearranged to be a vector, which is fed into the AOG-LSTM to predict the subbranch selection of the first or-node ... Note that the AOG is updated based on the annotated and predicted selection during training and test stages, respectively”; It can be seen that by utilizing the AOG-LSTM which generates a suggested path, as discussed above, to predict the subbranch of the first or-node, LIN teaches determining an edge connected to a front node based on the recommended path];
determining a child node connected to the edge based on the edge [Fig. 2, page 3, section 4.2, 2nd paragraph – page 4, Col. 1, 2nd paragraph, “which serves as the initial hidden state of the AOG-LSTM. It then encodes the initial AOG as an adjacency matrix. The matrix is further rearranged to be a vector, which is fed into the AOG-LSTM to predict the subbranch selection of the first or-node; It can be seen that utilizing the neural network to generate a suggested path to predict subbranch selection of the first or-node is reasonably understood to be included in determining a child node connected to the edge based on the edge];
As per claim 9, Lin and Tremblay teach the method of claim 8.
Lin further teaches
determining a recommended action type among the plurality of task actions based on the recommended path [Figs. 1-2, page 3, section 4.2, 2nd paragraph, “which serves as the initial hidden state of the AOG-LSTM. It then encodes the initial AOG as an adjacency matrix. The matrix is further rearranged to be a vector, which is fed into the AOG-LSTM to predict the subbranch selection of the first or-node”; Page 20, lines 21-22 of the instant specification state "Here, the action name may be an action type" and page 21, lines 1-2 of the instant specification state "an action type corresponding to an edge may be or not be recommended from the neural network", but do not appear to explicitly define an action type among a plurality of task actions. As such, the broadest reasonable interpretation of an action type includes an edge corresponding to an action. As such, by determining the subbranch selection of the first or-node utilizing the prediction generated by AOG-LSTM, LIN teaches determining a recommended action type among the plurality of task actions based on the recommended path];
determining the edge based on the recommended action type [Figs. 1-2, page 3, section 4.2, 2nd paragraph, “The matrix is further rearranged to be a vector, which is fed into the AOG-LSTM to predict the subbranch selection of the first or-node"; As discussed above, determining an action type is reasonably understood to encompass determining an edge. By teaching a method for determining a subbranch in the task plan based on the trained neural network, LIN teaches determining the edge based on the recommended action type].
As per claim 11, Lin teaches an autonomous system comprising an autonomous robot or vehicle [Fig. 1, page 1, introduction, “Automatically predicting and executing a sequence of actions given a specific task would be one quite expected ability for intelligent robots”], the autonomous system comprising:
generate a search tree based on a plurality of task states of the task and a plurality of task actions for performing the task [Fig. 2, page 2, Col. 1, 2nd paragraph, “we present a two-stage training method by employing a temporal And-Or graph (AOG) representation [5, 6]. First, we define the AOG for task description, which hierarchically decomposes a task into atomic actions according to their temporal dependencies. In this semantic representation, an and-node represents the chronological decomposition of a task (or sub-task); an or-node represents the alternative ways to complete the certain task (or sub-task); leaf-nodes represent the pre-defined atomic actions. The AOG can thus contain all possible action sequences for each task”; Examiner interprets the chronological decomposition of a task or sub-task comprises a state, interprets the graph comprising the states connected by the atomic actions as a search tree based on the task states of the task and a plurality of task actions, and interprets decomposing a task into atomic actions to define the AOG as generating a search tree based on the task states of the task and a plurality of task actions, and in the introduction, Lin teaches the intelligent robots are capable of performing the actions], estimate a recommended path for an internal connection of the search tree by inputting the plurality of task states and the plurality of task actions to a neural network based on the search tree [Fig. 4, page 4, Col. 1, 2nd and 3rd paragraphs, “Specifically, we collect a small set of samples annotated with the selections of all or-nodes given a scene image for each task and define the cross-entropy objective function to train the AOG-LSTM. Once the AOG-LSTM is trained, we adopt it to predict the sub-branch selections for all the or-nodes in the And-Or graph given different scene images, and generate the corresponding action sequences”; The examiner notes that, as written, the limitation is directed to path finding for a subset of the search tree or a subset of the arbitrary task. Page 8, lines 4-8 of the instant specification recites "The neural network (or an artificial neural network) may include a statistical training algorithm that mimics a biological nerve in machine training and cognitive science. The neural network may refer to a general model that has the ability to solve a problem, where artificial neurons (nodes) forming the network through synaptic combinations change a connection strength of synapses through training", page 8, line 13-page 9, line 2 of the instant specification lists example neural networks (including "CNN", "RNN" and "LSTM"), and page 17, lines 14-15 of the instant specification recites "The neural network may include an input layer 510, a hidden layer 530 and an output layer 550", but the instant specification does not appear to explicitly define a neural network. As such, the broadest reasonable interpretation of neural network based on the search tree, as written and in light of the specification, includes LIN's LSTM network trained on samples drawn from the AOG. Predicting a sub- branch is reasonably understood to be encompassed by estimating a recommended path for an internal connection of the search tree], generate the task plan by determining a target path that reaches from an initial state of the task to a target state of the task based on the recommended path [Fig. 1 shows 2 different task plans each comprising a sequence of actions for completing the task “make tea”, where, an initial state S0 of the task may describe a first location of the robot, and the target state st is the state after taking an action “pour into, cup”; page 5, section 6.2.1, 3rd paragraph, "Then we evaluate the accuracy of generating the action sequences, i.e., whether the task is completed successfully. We define the sequence accuracy as the fraction of complete correct sequences with respect to all predicted sequences"; As discussed above, LIN teaches generating a task plan for a task utilizing a neural network which outputs predicted next actions, or recommended paths. Page 11, lines 9-11 of the instant specification recites "The processor 100 may generate the task plan by determining a target path that reaches from an initial state of the task to a target state of the task based on the recommended path", but does not appear to explicitly define a target path. As such, as written and in light of the specification, the broadest reasonable interpretation of generating the task plan by determining a target path that reaches from an initial state of the task to a target state of the task based on the recommended path includes assessing the accuracy of the recommended path in achieving the goal of the task], and transmit the task plan to the autonomous robot or vehicle [Fig. 1 shows 2 different task plans each comprising a sequence of actions (perform by a robot) for completing the task “make tea”. Introduction, 1st paragraph, “executing a sequence of actions given a specific task would be one quite expected ability for intelligent robots. For example, to complete the task of “make tea” under the scene shown in Figure 1, an agent needs to plan and successively execute a number of steps, e.g., “move to the tea box”, “grasp the tea box”, where, the agent includes "intelligent robots"; It can be seen that the task plan is transmitted to the robot for executing]; and
wherein the autonomous system is configured to perform, by the autonomous robot or vehicle, each of a plurality of tasks of the task plan associated with the recommended path to achieve the target state of the task [Fig. 1 shows 2 different task plans each comprising a sequence of actions (perform by a robot) for completing the task “make tea”. Introduction, 1st paragraph, “executing a sequence of actions given a specific task would be one quite expected ability for intelligent robots. For example, to complete the task of “make tea” under the scene shown in Figure 1, an agent needs to plan and successively execute a number of steps, e.g., “move to the tea box”, “grasp the tea box”. In this paper we aim to train a neural network model to enable such a capability; where, the agent includes "intelligent robots"],
generate a trained neural network by training the neural network based on the plurality of task states and the plurality of task actions [page 1, introduction, 1st paragraph, “to complete the task of “make tea” under the scene shown in Figure 1, an agent needs to plan and successively execute a number of steps, e.g., “move to the tea box”, “grasp the tea box”. In this paper we aim to train a neural network model to enable such a capability”; page 2, Col. 1, 2nd paragraph, “we train an auxiliary LSTM network (named AOG-LSTM) to predict the selection at the or-nodes in the AOG, and can thus produce a large number of new valid samples (i.e., task-oriented action sequences) that can be used for training the Action-LSTM. Notably, training the AOG-LSTM requires only a few manually annotated samples (i.e., scene images and the corresponding action sequences)”; It can be seen that Scene images and corresponding action sequences are reasonably understood to include a plurality of task states and a plurality of task actions], and estimate the recommended path based on the trained neural network [page 6, section 7, 1st paragraph, “In this paper, we address a challenging problem, i.e., predicting a sequence of actions to accomplish a specific task under a certain scene, by developing a recurrent LSTM neural network”; It can be seen that predicting a sequence of actions to accomplish a specific task is reasonably understood to include estimating a recommended path], and
generate sequence data based on the plurality of task states, the plurality of task actions, and the task [page 6, section 7, 1st paragraph, “In this paper, we address a challenging problem, i.e., predicting a sequence of actions to accomplish a specific task under a certain scene, by developing a recurrent LSTM neural network”; Fig. 1 shows 2 different task plans each comprising a sequence of actions (sequence data) for completing the task “make tea”; It can be seen that the 2 task plans are generated based on the states such as the first location of a robot before taking any action from a list of actions such as: “move to, grasp, open, etc., and the actions take to complete the “make tea” task];
Lin does not explicitly teach
a processor;
the processor configured to;
a memory configured to store instructions executable by the processor;
acquire context present in the sequence data, determine output sequence data based on the context, and the sequence data, and train the neural network based on the output sequence data.
Tremblay teaches
a processor [Fig. 1, paragraph 0023, “at least one processor”];
the processor configured to [paragraph 0077, “the device includes at least one processor 702 for executing instructions that can be stored in a memory device”];
a memory configured to store instructions executable by the processor [Fig. 1, paragraph 0023, “memory 110 for including non-transitory computer-readable instructions for execution by the processor”];
acquire context present in the sequence data [abstract, “receive data involving the performance of a task in the physical world. The data can be provided as input to a perception network to infer a set of percepts about the task … The percepts can be provided as input to a plan generation network, which can infer a set of actions as part of a plan … The plan can be reviewed and any corrections made … Once the plan is verified as correct, the plan (and any related data) can be provided as input to an execution network that can infer instructions to cause the robot, and/or another robot, to perform the task”; paragraph 0026, “Once a plan is generated … The user can view the plan to determine whether the plan accurately represents the task to be performed”; Lin Fig. 1 shows 2 different task plans each comprising a sequence of actions (sequence data) for completing the task “make tea”, while Tremblay teaches once the task plan is generated, the plan is viewed to determine if it accurately represents the task to be performed. It can be seen that the process of viewing the plan to determine if it accurately represents the task to be performed, is to view the context/content of the plan/actions result to determining if the task is accurately performed based on the plan/actions, therefore, the combination of Lin and Tremblay teaches the above limitation], determine output sequence data based on the context, and the sequence data [paragraph 0026, “Once a plan is generated … The user can view the plan to determine whether the plan accurately represents the task to be performed. If not, the user can instruct the robot to capture data for another performance of the task using the objects 120, or can manually correct the plan to correctly represent the task to be performed, among other such options”; It can be seen that the generated plan which comprises task actions is viewed and corrected if needed to generate a corrected plan comprises task actions to accurately perform the task], and train the neural network based on the output sequence data [paragraph 0026, “Once a plan is generated … The user can view the plan to determine whether the plan accurately represents the task to be performed … After the plan is verified, the plan and/or associated data can be processed using an execution neural network, for example, to generate a set of instructions executable by the robot 102 to perform the task”].
It would have been obvious to one of ordinary skill in the art before the effective filing date of the invention to have modified the method of predicting a sequence of actions to complete a task of Lin to include the processes of acquiring context present in the sequence data, determining output sequence data, and training the neural network based on the output sequence data of Tremblay. Doing so would help predicting, by the neural network, a set of instructions executable by the robot to perform the task (Tremblay, 0026).
As per claim 12, Lin and Tremblay teach the autonomous system of claim 11.
Lin further teaches
generate nodes corresponding to the plurality of task states [page 3, Fig. 2(a)
PNG
media_image1.png
220
372
media_image1.png
Greyscale
It can be seen that the “and” nodes, “or” nodes, and leaf nodes of LI N's parse tree are reasonably understood to be encompassed by nodes corresponding to the plurality of task states]
generate the search tree by connecting the nodes via edges corresponding to the plurality of task actions [Figs. 1-2, page 2, Col. 1, 2nd paragraph, “In this semantic representation, an and-node represents the chronological decomposition of a task (or sub-task); an or-node represents the alternative ways to complete the certain task (or sub-task); leaf-nodes represent the pre-defined atomic actions”; It can be seen that alternate ways to complete the certain task and pre-defined atomic actions are reasonably understood to include actions].
As per claim 18, Lin and Tremblay teach the autonomous system of claim 11.
Lin further teaches
determine an edge connected to a front node in the search tree based on the recommended path [Fig. 2, page 3, section 4.2, 2nd paragraph – page 4, Col. 1, 2nd paragraph, “which serves as the initial hidden state of the AOG-LSTM. It then encodes the initial AOG as an adjacency matrix. The matrix is further rearranged to be a vector, which is fed into the AOG-LSTM to predict the subbranch selection of the first or-node ... Note that the AOG is updated based on the annotated and predicted selection during training and test stages, respectively”; It can be seen that by utilizing the AOG-LSTM which generates a suggested path, as discussed above, to predict the subbranch of the first or-node, LIN teaches determining an edge connected to a front node based on the recommended path];
determine a child node connected to the edge based on the edge [Fig. 2, page 3, section 4.2, 2nd paragraph – page 4, Col. 1, 2nd paragraph, “which serves as the initial hidden state of the AOG-LSTM. It then encodes the initial AOG as an adjacency matrix. The matrix is further rearranged to be a vector, which is fed into the AOG-LSTM to predict the subbranch selection of the first or-node; It can be seen that utilizing the neural network to generate a suggested path to predict subbranch selection of the first or-node is reasonably understood to be included in determining a child node connected to the edge based on the edge];
As per claim 19, Lin and Tremblay teach the autonomous system of claim 18.
Lin further teaches
determine a recommended action type among the plurality of task actions based on the recommended path [Figs. 1-2, page 3, section 4.2, 2nd paragraph, “which serves as the initial hidden state of the AOG-LSTM. It then encodes the initial AOG as an adjacency matrix. The matrix is further rearranged to be a vector, which is fed into the AOG-LSTM to predict the subbranch selection of the first or-node”; Page 20, lines 21-22 of the instant specification state "Here, the action name may be an action type" and page 21, lines 1-2 of the instant specification state "an action type corresponding to an edge may be or not be recommended from the neural network", but do not appear to explicitly define an action type among a plurality of task actions. As such, the broadest reasonable interpretation of an action type includes an edge corresponding to an action. As such, by determining the subbranch selection of the first or-node utilizing the prediction generated by AOG-LSTM, LIN teaches determining a recommended action type among the plurality of task actions based on the recommended path];
determine the edge based on the recommended action type [Figs. 1-2, page 3, section 4.2, 2nd paragraph, “The matrix is further rearranged to be a vector, which is fed into the AOG-LSTM to predict the subbranch selection of the first or-node"; As discussed above, determining an action type is reasonably understood to encompass determining an edge. By teaching a method for determining a subbranch in the task plan based on the trained neural network, LIN teaches determining the edge based on the recommended action type].
Claims 6-7 and 16-17 are rejected under 35 U.S.C. 103 as being unpatentable over Lin et al. in view of Tremblay et al. and further in view of Wartnick (US Patent 9,792,306).
As per claim 6, Lin and Tremblay teach the method of claim 1.
Lin further teaches
generating an information vector by encoding a task action and a task of the sequence data [Fig. 3
PNG
media_image2.png
198
374
media_image2.png
Greyscale
page 18, lines 15-16 of the instant specification recites "generate an information vector by encoding the task action and the task of the sequence data", but does not appear to explicitly define encoding the task action. Teaching a method which encodes the task and the associated AOG as a vector, LIN's teaching is encompassed by the broadest reasonable interpretation of generating an information vector by encoding a task action and a task of the sequence data];
generating the training data based on the hash code and the information vector [page 3, section 4.2, 1st – 2nd paragraphs, LIN: "In addition to capturing the task semantics, the AOG representation enables to generate large amount of valid samples (i.e., action sequences extracted from the AOG), which are significant for the recurrent neural network learning ... As illustrated in Figure 3, our model first extracts the features of the given scene image and the task, and maps them to a feature vector, which serves as the initial hidden state of the AOG-LSTM. It then encodes the initial AOG as an adjacency matrix. The matrix is further re-arranged to be a vector, which is fed into the AOG-LSTM to predict the subbranch selection of the first or-node ... Thus, we can train the AOG-LSTM only using a small number of annotated samples"; It can be seen that data which is significant for the recurrent neural network learning is reasonably understood to be encompassed by training data. The examiner notes that the sequences extracted from the AOG includes the task action of the task].
Lin and Tremblay do not teach
acquiring a hash code by performing a hash operation on a task state of the sequence data and the hash code.
Wartnick teaches
acquiring a hash code by performing a hash operation on a task state of the sequence data [Col. 6, lines 22-28, “Deduplication module 142 is configured to generate a fingerprint, or signature, for one or more segments of data. A fingerprint is an identifier of a respective segment stored in deduplicated data store 140. A fingerprint can be a checksum, hash value, or other such value that is calculated based upon data within the segment (e.g., within a file segment of client data)”; The examiner notes that files, or segments of files, may be reasonably understood to be encompassed by sequential data] and the hash code [Col. 9, lines 62-65, "In one embodiment, generating a finger-print involves using an algorithm to hash the segment of data"; The examiner notes that an algorithm which is used to hash a segment of data is reasonably understood to output a hash code].
It would have been obvious to one of ordinary skill in the art before the effective filing date of the invention to have modified the method of predicting a sequence of actions to complete a task of Lin to include the process of acquiring a hash code by performing a hash operation on a task state of the sequence data of Wartnick. Doing so would help transferring a smaller amount of data than if the data segments themselves were transmitted (Wartnick, Col. 4, lines 38-40).
As per claim 7, Lin, Tremblay and Wartnick teach the method of claim 6.
Lin further teaches
the generating of the information vector comprises acquiring a one-hot vector as the information vector by performing one-hot encoding on the task action and the task [page 5, Col. 1, 2nd paragraph, “fAt is a one-hot vector denoting a specific task. fT is a concatenated by two one-hot vectors, denoting the primitive action and object, respectively”; As discussed with reference to figure 3, the vectors generated are utilized by the neural network].
As per claim 16, Lin and Tremblay teach the autonomous system of claim 11.
Lin further teaches
generate an information vector by encoding a task action and a task of the sequence data [Fig. 3
PNG
media_image2.png
198
374
media_image2.png
Greyscale
page 18, lines 15-16 of the instant specification recites "generate an information vector by encoding the task action and the task of the sequence data", but does not appear to explicitly define encoding the task action. Teaching a method which encodes the task and the associated AOG as a vector, LIN's teaching is encompassed by the broadest reasonable interpretation of generating an information vector by encoding a task action and a task of the sequence data];
generate the training data based on the hash code and the information vector [page 3, section 4.2, 1st – 2nd paragraphs, LIN: "In addition to capturing the task semantics, the AOG representation enables to generate large amount of valid samples (i.e., action sequences extracted from the AOG), which are significant for the recurrent neural network learning ... As illustrated in Figure 3, our model first extracts the features of the given scene image and the task, and maps them to a feature vector, which serves as the initial hidden state of the AOG-LSTM. It then encodes the initial AOG as an adjacency matrix. The matrix is further re-arranged to be a vector, which is fed into the AOG-LSTM to predict the subbranch selection of the first or-node ... Thus, we can train the AOG-LSTM only using a small number of annotated samples"; It can be seen that data which is significant for the recurrent neural network learning is reasonably understood to be encompassed by training data. The examiner notes that the sequences extracted from the AOG includes the task action of the task].
Lin and Tremblay do not teach
acquire a hash code by performing a hash operation on a task state of the sequence data and the hash code.
Wartnick teaches
acquire a hash code by performing a hash operation on a task state of the sequence data [Col. 6, lines 22-28, “Deduplication module 142 is configured to generate a fingerprint, or signature, for one or more segments of data. A fingerprint is an identifier of a respective segment stored in deduplicated data store 140. A fingerprint can be a checksum, hash value, or other such value that is calculated based upon data within the segment (e.g., within a file segment of client data)”; The examiner notes that files, or segments of files, may be reasonably understood to be encompassed by sequential data] and the hash code [Col. 9, lines 62-65, "In one embodiment, generating a finger-print involves using an algorithm to hash the segment of data"; The examiner notes that an algorithm which is used to hash a segment of data is reasonably understood to output a hash code].
It would have been obvious to one of ordinary skill in the art before the effective filing date of the invention to have modified the method of predicting a sequence of actions to complete a task of Lin to include the process of acquiring a hash code by performing a hash operation on a task state of the sequence data of Wartnick. Doing so would help transferring a smaller amount of data than if the data segments themselves were transmitted (Wartnick, Col. 4, lines 38-40).
As per claim 17, Lin, Tremblay and Wartnick teach the autonomous system of claim 16.
Lin further teaches
acquire a one-hot vector as the information vector by performing one-hot encoding on the task action and the task [page 5, Col. 1, 2nd paragraph, “fAt is a one-hot vector denoting a specific task. fT is a concatenated by two one-hot vectors, denoting the primitive action and object, respectively”; As discussed with reference to figure 3, the vectors generated are utilized by the neural network].
Prior Art
The prior art made of record and not relied upon is considered pertinent to applicant’s disclosure.
Kuo et al. (US Patent 12,304,081) describes task and motion planning for a robot.
Keselman et al. (US Pub. 2020/0097015) describes a method for motion planning of an autonomous driving machine.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to TRI T NGUYEN whose telephone number is 571-272-0103. The examiner can normally be reached M-F, 8 AM-5 PM, (CT).
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, OMAR FERNANDEZ can be reached at 571-272-2589. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/TRI T NGUYEN/Examiner, Art Unit 2128
/OMAR F FERNANDEZ RIVAS/Supervisory Patent Examiner, Art Unit 2128
Prosecution Insights