CTNF 18/910,587 CTNF 76255 DETAILED ACTION 07-03-aia AIA 15-10-aia The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA. Specification 07-29 AIA The disclosure is objected to because of the following informalities: In ¶[0027], “respective scores the set of” appears that is should be “respective scores of the set of”. In ¶[0055], “Unfortanatley” should be “Unfortunately”. In ¶[0058], “can covert” should be “can convert”. In ¶[0058], “can encode the encodes the current state” appears that it should be “can encode the current state” or “encodes the current state:. In ¶[0077], “appliacble” should be “applicable”. In ¶[0077], “enoded” should be “encoded”. In ¶[0077], “represntations” should be “representations”. In ¶[0077], “gererate” should be “generate”. In ¶[0077], “priritized” should be “prioritized”. In ¶[0081], “In the blocks domain, The blocks domain” should be “The blocks domain”. In ¶[0081], “pick up b2” appears that it should be “pick up b3” as illustrated in Figure 3. In ¶[0083], “Similiarly” should be “Similarly”. In ¶[0083], “Additonally” should be “Additionally”. In ¶[0083], “referncing” should be “referencing”. In ¶[0084], “instnace” should be “instance”. In ¶[0084], “compris” should be “comprise”. In ¶[0085], “instnacnes” should be “instances”. In ¶[0087], “separatley” should be “separately”. In ¶[0087], “infrencing” should be “inferencing”. In ¶[0088], “in paticular” should be “in particular”. In ¶[0088], “(e.g., the encoded representations 420 and the encoded representations 422)” appears that it should be “(e.g., the encoded representations 420 and the encoded representations 424)”. See Figure 4. In ¶[0095], “(e.g., receive from retrieval pool 128)” appears that it should be “(e.g., received from retrieval pool 128)”. In ¶[0146], “according to as ‘as a service’ technology” appears that it should be “according to an ‘as a service’ technology” . Appropriate correction is required. Claim Objections 07-29-01 AIA Claim s 4, 15, and 20 are objected to because of the following informalities: These claims set forth a limitation of determining using a similarity function “respective scores the set of applicable actions” which appears that it should be “respective scores of the set of applicable actions” . Appropriate correction is required. Claim Rejections - 35 USC § 103 07-20-aia AIA 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. 07-21-aia AIA Claim s 1 to 2, 11, 13 to 14, and 19 are rejected under 35 U.S.C. 103 as being unpatentable over Meuleau et al . (U.S. Patent Publication 2020/0122039) in view of Das et al . (U.S. Patent Publication 2025/0053156) . Concerning independent claims 1, 13, and 19, Meuleau et al . discloses a system, method, and non-transitory computer-readable medium, comprising: “a memory that stores computer executable components, and a processor that executes at least one of the computer executable components that:” – systems and methods include a behavior generation device 102 and a processing device including one or more central processing units 104 having access to memory 106 to retrieve instructions stored thereon to perform a series of tasks described (¶[0043]: Figure 1); “receives a current state, a set of applicable actions, and a set of goal states that define a planning problem” – behavior control module 116 keeps track of one or more goals for an AI agent; behavior planning module 118 solves problems by determining a plan that includes a sequence of actions to be carried out by control module 116, e.g. , actions applied to world objects, that will change the planning state in an attempt to satisfy the goals; behavior planning module 118 receives as inputs a first state of the world model at a first time, e.g. , the current state and a goal for an agent (¶[0037] - ¶[0038]: Figure 1); a state-action pair includes data describing a state and one or more associated actions, e.g. , {current state, action 1}, {current state, action 2}; planning module 118 uses estimate 344 to efficiently evaluate state-action pairs when exploring possible future states and actions available in the future states (¶[0043]: Figure 3). Concerning independent claims 1, 13, and 19, Meuleau et al . discloses a machine learning model to generate and evaluate a plan to perform a goal based on a current state and a set of applicable actions. Additionally, Meuleau et al . discloses state descriptions 322 and action descriptions 320 are parsed into a ML encoded state description 328 and a ML encoded action description 320. (¶[0044]: Figure 3) Here, a ML encoded state description 328 and a ML encoded action description 320 are “encoded representations of the current state, the set of applicable actions”. However, Meuleau et al . does not generate a plan “via a language model” and does not evaluate a plan “via a graph search algorithm . . . based on the encoded representations.” Concerning independent claims 1, 13, and 19, Das et al . teaches efficiently planning a sequence of actions utilizing commonsense knowledge from a large language model. (Abstract) Specifically, Das et al . teaches: “generates, via a language model, a plan for the planning problem, wherein generating the plan for the planning problem comprises:” – a search network utilizes commonsense knowledge from large language models to find unseen objects; robotic agent utilizes a visual input to efficiently plan a sequence of actions for simultaneous object search and rearrangement (Abstract); efficient planning makes rearrangement more effective by optimizing the sequence of actions and minimizing the time and effort required to achieve the goal state (¶[0004]); a large language model (LLM) based search network is triggered to predict a probable receptable for each unseen object (¶[0008]); “generating, via an encoder, respective encoded representations of the current state, the set of applicable actions” – a graph embedding is created from the spatial graph representation of the untidy current state and the user-specified tidy goal state via a graph representation network (GRN) (“encoded representations of the current state”) (¶[0009]); Deep RL (P-DQN) network with hybrid action space is used to plan action sequences for simultaneous object search and rearrangement by resolving blocked goal and the swap cases (¶[0049]; Figure 2: Step 210); a graph embedding is created from a spatial graph representation of the untidy current state and the user-specified tidy goal state via the graph neural network (GNN); the graph embedding enables state space to understand semantic and geometric information of the untidy current state and user-specified tidy goal states (¶[0060] - ¶[0061]: Figure 2); a RoBERTa-Large model known in the art is used to generate pairwise embeddings (E VR ) for {W i V)} i=1,2, . . . ,N p . and {W i R }i=1,2, . . . , N R in the current state (¶[0072]); a new Graph Representation Network (GRN) with the encoder-decoder (“via an encoder”) is used to generate meaningful embeddings from G C and G g for Deep RL state space to incorporate the residual relative path length notion between every pair of current and goal state nodes (¶[0074]); each action {a i =(k,p k )} in the sequence of actions {a i } = 1,2, . . . , K of length K is made up of a discrete action k, denoting the index of the selected object or the probable receptacle, followed by a continuous parameter P k which signifies the location for object placement or receptacle search (“respective encoded representations of . . . the set of applicable actions”) (¶[0075]); “evaluating, via a graph search algorithm, a set of plans based on the encoded representations” – a graph-based state representation produces a scalable and effective planner to minimize the number of steps taken and to resolve blocked goal and swap cases (Abstract); unique graph-based state representation produces a scalable and effective planner that interleaves object search and rearrangement to minimize the number of steps taken and overall traversal of the agent, as well as to resolve blocked goal and swap cases (¶[0035]); a Deep RL state space is defined with a novel graph-based state representation for the current and the goal state that incorporates geometric information about object; this representation compactly encodes the scene geometry that aids in rearrangement planning (¶[0050]); a spatial graph (G={V,E}) representation of the current state (refers to untidy current state) and the goal state (refers to user-specified tidy goal state) is created (¶[0074]). Concerning independent claims 1, 13, and 19, Das et al ., then, teaches generating a plan via a large language model with embeddings from an encoder of a current state to a goal state using actions and evaluating a plan with a graph neural network (GNN). An objective is to provide efficient task planning under partial observability of a room in a room arrangement task. (¶[0004]) It would have been obvious to one having ordinary skill in the art to apply a large language model and a graph search algorithm as taught by Das et al . to generate a plan from encoded state descriptions and encoded action descriptions to a goal state in Meuleau et al . for a purpose of providing efficient task planning under partial observability of a room in a room arrangement task. Concerning claims 2 and 14, Meuleau et al . discloses that ML encoded state and ML encoded action are processed with a recurrent neural network (RNN) to generate an estimate of a value of the state description and the action description (¶[0014]); there are a plurality of state descriptions 322 and action descriptors 320 sent from planning module 118 to ML module 120; a state descriptor 322 is matched with an action description 320 in a state-action pair; state descriptions 322 and action descriptions 320 represent future states and future actions evaluated by planning module 118; state descriptions 322 and action descriptions 320 are parsed into a ML encoded state description 328 and a ML encoded action description 320 (“respective encoded representations of the set of applicable actions separate from the respective encoded representations of the current state and the set of goal states”) (¶[0044]: Figure 3). Here, Applicants’ limitation of “a bi-encoder” only appears to imply that an action is encoded separately from a state as described at ¶[0087] of the Specification. Concerning claim 11, Meuleau et al . discloses that a plan that includes a plurality of sequential actions for an agent is generated; a plurality of sequential actions is chosen based on at least the value estimate (“selects an action from a retrieval pool that comprises the set of applicable actions”) (¶[0014]); a behavior planner is a problem solving module that requires a goal to compute a plan to drive the system from the current state to a goal state; a plan includes a sequence of actions or a sequence of sets of actions that can be executed simultaneously (¶[0025]); a behavior planner is a problem solving module that requires a goal to compute a plan to drive the system from the current state to a goal state; a plan includes a sequence of actions or a sequence of sets of actions that can be executed simultaneously (¶[0038]: Figure 1); a state-action pair includes data describing a state and one or more associated actions ( e.g., {current state, action 1}, {current state, action 2}, and the like) (“selects an action from the retrieval pool that comprises the set of applicable actions”); planning module 118 uses estimate 344 to efficiently evaluate ( e.g., including ranking, choosing and eliminating) state-action pairs when exploring possible future states and actions available in the future states (¶[0043]: Figure 3) . 07-22-aia AIA Claim 3 is rejected under 35 U.S.C. 103 as being unpatentable over Meuleau et al . (U.S. Patent Publication 2020/0122039) in view of Das et al . (U.S. Patent Publication 2025/0053156) as applied to claim 1 above, and further in view of Li et al . (U.S. Patent Publication 2023/0169110) . Meuleau et al . discloses encoded representations of a set of applicable actions, but does not expressly disclose generating these encoded representations “in an offline process and stores” the respective encoded representations of the set of applicable actions “for subsequent use”. However, Li et al . teaches analogous art of content retrieval that encodes textual data into text embedding representations via a machine learning model for selecting matching visual assets. (Abstract) Specifically, Li et al . teaches an offline phase involves use of a trained representation model to process asset libraries to convert assets to embedding vectors which are then stored in an asset index. (¶[0050]) Li et al ., then, teaches an offline phrase that processes assets that are converted into embedding vectors and then stored so as to provide “in an offline process and stores” the set of assets “for subsequent use.” An objective is to provide an improved system and method for locating and retrieving content. (¶[0003]) It would have been obvious to one having ordinary skill in the art to provide encoded representations of a set of applicable actions in Meuleau et al . that are converted into embedding vectors and stored in an offline process as taught by Li et al . for a purpose of obtaining an improved system and method for locating and retrieving content . 07-22-aia AIA Claim s 4 and 15 are rejected under 35 U.S.C. 103 as being unpatentable over Meuleau et al . (U.S. Patent Publication 2020/0122039) in view of Das et al . (U.S. Patent Publication 2025/0053156) as applied to claim s 1 and 13 above, and further in view of Tao et al . (U.S. Patent Publication 2025/0245516) . Meuleau et al . discloses that ML encoded state and ML encoded action are processed with a recurrent neural network to generate an estimate of a value of the state description and the action description, a plan that includes a plurality of sequential actions for an agent is generated, and a plurality of sequential actions is chosen based on at least the value estimate. (Abstract) Input 344 from the machine learning module 120 is a value estimate for a state-action pair, and planning module 118 uses the estimate 344 to efficiently evaluate ( e.g., including ranking, choosing and eliminating) state-action pairs when exploring possible future states. (¶[0043]: Figure 3) Meuleau et al ., then, discloses ranking actions based on an estimate of a value, but does not disclose that an action is selected by determining “respective scores” and “using a similarity function . . . based on a similarity to the current state of the set of goal states.” However, Tao et al . teaches systems and methods for models based reward design with an image encoder to generate image-based embeddings of a current state of a vehicle and passing a goal of an autonomous vehicle through a text encoder to generate text-based embeddings of the goal. A similarity score is determined representing a similarity between the image-based embeddings of the current state and the text-based embeddings of the goal. An action policy corresponding to a control of the vehicle is optimized based on the reward function. (Abstract; ¶[0005]) Tao et al ., then, teaches “determining . . using a similarity function, respective scores the set of applicable actions based on a similarity to the current state . . . .” An objective is to perform reinforcement learning to learn a policy mapping of states to actions to enhance a vehicle’s ability to navigate safely. (¶[0004]) It would have been obvious to one having ordinary skill in the art to determine a similarity score using a similarity function as taught by Tao et al . to rank actions in Meuleau et al . for a purpose of performing reinforcement learning to learn a policy of mapping states to actions . Allowable Subject Matter 12-151-08 AIA 07-43 12-51-08 Claim s 5 to 10, 12, 16 to 18, and 20 are objected to as being dependent upon a rejected base claim, but would be allowable if rewritten in independent form including all of the limitations of the base claim and any intervening claims. Conclusion 07-96 The prior art made of record and not relied upon is considered pertinent to Applicants’ disclosure. Lu et al., Anderson et al., Manikonda et al., Schaul et al., Dvorak et al., Baldua et al., Rossi et al., and McMorran et al. disclose related prior art. Any inquiry concerning this communication or earlier communications from the examiner should be directed to MARTIN LERNER whose telephone number is (571) 272-7608. The examiner can normally be reached Monday-Thursday 8:30 AM-6:00 PM. If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Richemond Dorvil can be reached at (571) 272-7602. 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. /MARTIN LERNER/Primary Examiner Art Unit 2658 June 2, 2026 Application/Control Number: 18/910,587 Page 2 Art Unit: 2658 Application/Control Number: 18/910,587 Page 3 Art Unit: 2658 Application/Control Number: 18/910,587 Page 4 Art Unit: 2658 Application/Control Number: 18/910,587 Page 5 Art Unit: 2658 Application/Control Number: 18/910,587 Page 6 Art Unit: 2658 Application/Control Number: 18/910,587 Page 7 Art Unit: 2658 Application/Control Number: 18/910,587 Page 8 Art Unit: 2658 Application/Control Number: 18/910,587 Page 9 Art Unit: 2658 Application/Control Number: 18/910,587 Page 10 Art Unit: 2658 Application/Control Number: 18/910,587 Page 11 Art Unit: 2658
Prosecution Insights