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 .
Status of Claims
Claims 1-3, 5-12, 14-18, and 21 were rejected in the Non-Final Office action mailed on 11/28/2025. Applicant’s amended claimset, entered on 03/27/2026, amended Claims 1, 10, and 21, cancelled Claim 5, and added new Claims 22-23. Herein this Final Office Action, Claims 1-3, 6-12, 14-18, and 21-23 are rejected.
Response to Arguments
Applicant’s arguments filed 03/27/2026, with respect to Rejections under 35 U.S.C. 101 for Claims 1-3, 6-12, 14-18, and 21, have been fully considered and are not persuasive.
On Pages 1-3, Applicant argues “As amended, claim 1 recites the added limitations of "following fulfillment of the batch by the picker, computing a batched delivery cost based on a cost associated with fulfillment of the batch and a time to claim the batch from the unclaimed order pool;" and "retraining the machine-learned counterfactual cost model based on the batched delivery cost." These limitations cannot be characterized as directed to any abstract idea. . . These steps relate to computing metrics associated with fulfillment of the batched order, which is then used to retrain the machine-learning counterfactual model. Though machine-learning models are understandably based in and rely on mathematical principles, the recitation of training or deploying such models are not themselves math or related computations. As such, these steps cannot be characterized as reciting any mathematical concepts. Retraining of a machine-learning model is a technical step performed by computers. It is not innately an economic or interaction in any way shape or form. To be clear, the step of training the machine-learning model has no human parties seeking to transact, interact, or otherwise economize. As such, these limitations cannot be characterized as reciting any excluded method of organizing human activity. As noted above, the training of a machine-learning model is technical in nature and only feasibly-performed by a computing device, not a human mind. The human mind, though it served as inspiration to the field of machine learning, is not itself architected to operate on machine-learning models. The human mind does not input computer-readable data, and does not operate on this data through functions and learned parameters. In sum, these limitations cannot be characterized as a mental processes abstract idea.” Examiner does not fully agree.
Examiner agrees, in part, that training, or retraining a machine learning model is not an abstract idea. However, computing a cost based on certain information and updating a mathematical model based on the computed cost is a recited abstract idea as discussed in greater detail in the rejection section below. Using machine learning to update the model by means of retraining is merely using computer components in their ordinary capacity as a tool to apply the abstract idea per MPEP 2106.05(f).
On pages 3-5, Applicant argues that the recited abstract idea is integrated into a practical application by providing an improvement in the functioning of a computer or other technology under MPEP 2106.04(d). Applicant argues “The additional elements embody an improvement to the field of machine-learning. The additional elements specify a particular architecture to the machine-learning model using two different models trained to predict metrics related to alternative outcomes. The outputs are then weighed against one another for the system to time the release of an order to an unclaimed pool. Contextually, batching is a technical problem in the field. Batching has upsides and downsides. The balance between holding an order, for the potential of batching, versus releasing an order to an unclaimed pool comes down to timing, wherein timing affects downstream efficiency of the system. "However, it is difficult to assess when to release orders for fulfillment most efficiently. For example, immediately releasing a new customer order to be claimed by assistants results in fast fulfillment, but delaying releasing of the order until it can be combined ( or "batched") with other customer orders can reduce average order fulfillment costs by allowing a single assistant to fulfill all the orders in the batch in a single trip, reducing costs for the customers .... However, it is difficult to determine how best to balance the tradeoff because predicting the relevant metrics cannot be done mentally due to the complexity." Specification ¶ [0002]. This technical problem is solved through the deployment of the two-model architecture. As each model is precisely trained to predict one subtask of the prediction problem, the architecture embodies the advantage in deployment of the two models in parallel. "In some embodiments, a training module 311 trains the present cost model 305 and/or the future cost model 310 based on their respective features. The training features may be obtained from the logged data 312, either directly, or as a result of additional preprocessing. The training module 311 may also retrain the models 305, 310 based on the additional data obtained from prior decisions of the batching module 250 about how to batch orders and the actual data logged in association with the fulfillment of those batched orders, such as their actual delivery cost, actual time of arrival, whether they were late, etc." Specification ¶ [0060]. Moreover, with the actual fulfillment data captured following the batching decision, the methodology implements a retraining step for tuning the machine-learned counterfactual model. The additional elements particularly outline the computation of the actual batched delivery cost for the batched order to use as a supervision label for retraining of the machine-learned counterfactual model. The effect is a model that can be iteratively tuned, improving its precision over time, order-by-order. In sum, the recited technological solution thereby integrates the alleged judicial exception into a practical application-supporting a finding of eligibility under Step 2A, Prong Two.” (Emphasis added). Examiner does not agree.
Examiner responds that “batching” is a commercial (i.e. business) problem, not a technical problem. The balance between releasing an order, delaying the release, and batching an order are business decision that affects the efficiency of business operations, i.e. optimal allocation of commercial assets (e.g. orders and pickers). The “tradeoff between fulfillment speed and fulfillment cost” discussed in Specification ¶2 is a business decision which is solved using mathematical calculations, i.e. models. Thus, advantages in “better” allocation of commercial assets based on “better” business decisions would be an improvement in the abstract idea itself, not a patent eligible improvement in technology under MPEP 2106.05(a).
Moving to the machine learning, Specification ¶48 states “The machine-learning training module 230 trains machine-learning models used by the online concierge system 140. The online concierge system 140 may use machine-learning models to perform functionalities described herein. Example machine-learning models include regression models, support vector machines, naïve bayes, decision trees, k nearest neighbors, random forest, boosting algorithms, k-means, and hierarchical clustering. The machine-learning models may also include neural networks, such as perceptrons, multilayer perceptrons, convolutional neural networks, recurrent neural networks, sequence-to-sequence models, generative adversarial networks, or transformers. A machine-learning model may include components relating to these different general categories of model, which may be sequenced, layered, or otherwise combined in various configurations.” Given the breadth of machine learning techniques that could be used, i.e. essentially any type of machine learning, per Specification ¶48, Examiner cannot conclude that the specification provides sufficient technical explanation for the claims to provide “an improvement to the field of machine-learning” under MPEP 2106.05(a). Instead, the claims use machine learning as a tool in its ordinary capacity to apply the recited abstract idea, i.e. modeling cost, batching orders, and determining timing of release or orders.
Specification ¶60 merely discusses the types of information input to and output from the machine leaning. Such feature is insufficient technical explanation under MPEP 2106.05(a). Similarly, PEG Example 47 Claim 2 shows that limiting the training data and output data of an “artificial neural network” to certain types of “continuous” information is insufficient to provide patent eligible subject.
On Pages 5-6, Applicant argues, Regarding Step 2B, Moreover, the additional elements are non-routine, unconventional, and not well understood activity in the technological field, thereby amounting to an inventive concept supporting eligibility under Step 2B. The MPEP provides that "an examiner should determine that an element ( or combination of elements) is well-understood, routine, conventional activity only when the examiner can readily conclude, based on their expertise in the art, that the element is widely prevalent or in common use in the relevant industry" (italics for emphasis). MPEP § 2106.05(d). The additional elements are not so widely prevalent. Wide prevalence requires a high degree of understanding among artisans in the field, such that it need not be described in detail. See id, further referencing 35 U.S.C. § 112(a). The additional elements are not so widely prevalent as to render them conventional, routing, or well understood. Critically, the prior art does not teach nor suggest the additional elements, specifically the dual-model architecture of the machine-learning paradigm. Accordingly, the additional elements further support finding of eligibility under Step 2B, in addition to Step 2A, Prong Two.” (Emphasis added). Examiner disagrees.
Examiner responds that, although MPEP 2106.05(d) was not a basis for the rejection, the specification, in light of 35 USC 112(a), does support a determination that the machine learning elements used are generic, e.g. “off-the-shelf,” computer components. As cited above, Specification ¶48 lists methods of machine learning, e.g. “neural network,” by name, without specific explanation as to how each of these methods would execute the training of a model. Therefore, in compliance with 35 U.S.C. 112(a), the disclosure supports the use of, and only the use of, common machine learning techniques known in the art.
Additionally, although the claims are not rejected under 35 USC 102 or 103, Applicant is reminded that “[t]he Supreme Court’s decisions make it clear that judicial exceptions need not be old or long-prevalent, and that even newly discovered or novel judicial exceptions are still exceptions.” MPEP 2106.04.I. Thus, the lack of art rejection does not show patent eligibility.
On Pages 16-17, Applicant argues that the claims recite patent eligible subject matter in the “Conclusion” section. Examiner does not agree. As discussed above and below, the rejection under 35 U.S.C. 101 remains.
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 1-3, 6-12, 14-18, and 21-23 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1
Claims 1-3, 6-9, and 22-23 recite a method (i.e. a process)., Claims 10-12 and 14-18 recite a non-transitory computer-readable medium (i.e. a machine or manufacture), and Claim 21 recites a non-transitory computer-readable medium (i.e. a machine or manufacture). Therefore, Claims 1-3, 6-12, 14-18, and 21-23 all fall within the one of the four statutory categories of invention of 35 U.S.C. 101.
Step 2A, Prong One
Independent Claim 1 recites the abstract idea of:
“maintaining an unclaimed order pool of orders received from users and available to be claimed by pickers for fulfillment on behalf of the users, and a batching candidate pool of orders not yet available to be claimed by pickers;
receiving, from a user, an order for delivery of a set of items to an address of the user;
adding the order to the batching candidate pool;
deriving, from the order, a set of features characterizing the order;
providing the set of features as input to a . . . present cost model to obtain a first estimate representing an estimated delivery cost if the order were released from the batching candidate pool to the unclaimed order pool without attempting batching, the . . . present cost model being . . . for operating on the set of features as input to the [model] and for outputting the first estimate representing the estimated delivery cost without attempting batching;
providing the set of features as input to a . . . counterfactual cost model to obtain a plurality of tuples, for candidate batching options, the . . . counterfactual cost model being . . . for operating on the set of features as input to the [model] and for outputting the plurality of tuples, each tuple comprising:
a possible batch size,
an estimated delivery cost if the order were released to the unclaimed order pool during a given future time window allowing for a possibility of batching the order with orders of other users, and
a probability associated with the estimated delivery cost;
determining, based at least in part on the first estimate and on the tuples, whether to delay release of the order to the unclaimed order pool in order to attempt batching of the order;
responsive to determining that the release of the order to the unclaimed order pool should be delayed, delaying release of the order to the unclaimed order pool for a period of time;
during the future time window following the period of time, receiving another order for delivery of another set of items to another address of another user; [[and]]
batching the other order for delivery along with the order and releasing the batch comprising the order and the other order to the unclaimed order pool;
receiving a request to claim the batch comprising the order and the other order;
transmitting information on the batch comprising the order and the other order to a . . . picker for fulfillment;
following fulfillment of the batch by the picker, computing a batched delivery cost based on a cost associated with fulfillment of the batch and a time to claim the batch from the unclaimed order pool; and
[updating] the. . . counterfactual cost model based on the batched delivery cost.”
The limitations stated above are processes/ functions that under broadest reasonable interpretation covers (1) maintaining a pool of orders available to pickers for fulfillment and a batching candidate pool of orders that are temporarily unavailable to pickers, (2) receiving an order that is added to the batching candidate pool, (3) deriving features of that order, (4) inputting the features into a model to obtain estimated delivery cost if the order were released without attempting batching, (5) inputting the features into a model to obtain tuples comprising certain data including batch size, estimated cost if the order were released, and probability associated with the estimated cost, (6) determining whether to delay the release of the order based on the estimate and the tuple, (7) delaying the release responsive to determining that the release should be delayed, (8) receiving a second order during the future time window and batching the orders, (9) releasing the batch, (10) receiving a request to claim the batch, (11) transmitting information on the batch to a picker, (12) after fulfilment, computing a cost based on certain information, and (13) updated the model based on the computed cost, all of which are mathematical calculations (i.e. using a model to calculate outputs including cost and probability, computing actual cost, and updating the parameters of the model), which are mathematical concepts, an abstract idea, under MPEP 2106.04(a)(2)I, and managing personal behavior by following rules and interacting between people (i.e. determining when to release orders to pickers are at least “following rules or instructions” and communication of information is at least a “social activity”) and commercial or legal interactions (i.e. receiving orders, maintaining pools of orders, batching certain orders based on order features, predicting business metrics based on batching, releasing a batch, receiving a request, communicating necessary information for fulfilment, and providing cost feedback after fulfilment are at least “marketing or sales activities or behaviors” or “business relations”), which are certain methods of organizing human activity, an abstract idea, under MPEP 2106.04(a)(2)II. The mere the recitation of generic computer components (i.e., the “computer system,” “processor,” “computer-readable medium,” and “machine learned [models],” “a multilayer perceptron comprising a plurality of layers,” “retraining the machine-learned [models],” and “client device”) implementing the identified abstract idea does not take the claim out of the mathematical concepts or certain methods of organizing human activity groupings. MPEP 2106.04(d). If a claim limitation, under its broadest reasonable interpretation, covers “mathematical calculations,” “managing personal behavior or relationships or interactions between people,” and “commercial or legal interactions” but for the recitation of generic computer components, then it falls in the mathematical concepts or certain methods of organizing human activity groupings of abstract ideas. MPEP 2106.04. Therefore, Claim 1 recites an abstract idea.
Step 2A, Prong Two
The judicial exception is not integrated into a practical application. Claim 1 as a whole amounts to: (i) merely invoking generic components as a tool to perform the abstract idea or “apply it” (or an equivalent) and (ii) generally links the use of a judicial exception to a particular technological environment or field of use. The claim recites the additional elements of:
(i) “computer system”
(ii) “processor,”
(iii) “computer-readable medium,”
(iv) “machine learned [models]” including
(v) “a multilayer perceptron comprising a plurality of layers” and
(vi) “retraining the machine-learned [models],” and
(vii) “client device.”
The additional elements of (i) “computer system” (Fig. 2 and ¶29 shows “online concierge system 140.”), (ii) “processor” (¶73 shows “a processor comprises one or more processors or processing units that, individually or together, perform the steps of instructions stored on a computer-readable medium”), (iii) “computer-readable medium” (¶73 shows “a computer-readable medium comprises one or more computer-readable media that, individually or together, comprise instructions that, when executed by one or more processors, cause the one or more processors to perform, individually or together, the steps of the instructions stored on the one or more computer-readable media.”), and (iv) “machine learned [models]” being (v) “a multilayer perceptron comprising a plurality of layers” (Fig. 2 and ¶48 shows “The machine-learning training module 230 trains machine-learning models used by the online concierge system 140. . . Example machine-learning models include regression models, support vector machines, naïve bayes, decision trees, k nearest neighbors, random forest, boosting algorithms, k-means, and hierarchical clustering. The machine-learning models may also include neural networks, such as perceptrons, multilayer perceptrons, convolutional neural networks, recurrent neural networks, sequence-to-sequence models, generative adversarial networks, or transformers. A machine-learning model may include components relating to these different general categories of model, which may be sequenced, layered, or otherwise combined in various configurations.” See also ¶75 further discussing machine learning model), (vi) “retraining the machine-learned [models]” (¶51 shows “The machine-learning training module 230 may apply an iterative process to train a machine-learning model whereby the machine-learning training module 230 updates parameter values of the machine-learning model based on each of the set of training examples. . . . The machine-learning training module 230 updates the set of parameters for the machine-learning model based on the score generated by the loss function. For example, the machine-learning training module 230 may apply gradient descent to update the set of parameters.” See also ¶¶48-51 further showing that the claimed training is does not require a specific method of training, but could be performed by any known method for training a machine learning model.), and (vii) “client device” (Fig. 1 and ¶18 shows “The picker client device 110 can be a personal or mobile computing device, such as a smartphone, a tablet, a laptop computer, or desktop computer.”), are recited at a high-level of generality, such that, when viewed as whole/ordered combination (Fig. 1-2, ¶11, ¶29, ¶48, ¶73, and ¶75 shows elements in combination.), they amount to no more than mere instruction to apply the judicial exception using generic computer components or “apply it” (See MPEP 2106.05(f)).
The (i) “computer system,” (ii) “processor,” (iii) “computer-readable medium,” (iv) “machine learned [models],” (v) “a multilayer perceptron comprising a plurality of layers,” (vi) “retraining the machine-learned [models],” and (vii) “client device,” when viewed as whole/ordered combination (Fig. 1-2, ¶11, ¶29, ¶48, ¶73, and ¶75 shows elements in combination.), does no more than generally link the use of the judicial exception to a particular technological environment or field of use (i.e. online computer environment) (See MPEP 2106.05(h)).
Accordingly, these additional elements, when viewed as a whole/ordered combination (Fig. 1-2, ¶11, ¶29, ¶48, ¶73, and ¶75 shows elements in combination.), do not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea. Thus, the claim is directed to an abstract idea.
Step 2B
As discussed above with respect to Step 2A Prong Two, the additional elements amount to no more than: (i) “apply it” (or an equivalent) and (ii) generally link the use of a judicial exception to a particular technological environment or field of use, and are not a practical application of the abstract idea. The same analysis applies here in Step 2B, i.e., (i) merely invoking the generic components as a tool to perform the abstract idea or “apply it” (See MPEP 2106.05(f)) and (ii) generally linking the use of a judicial exception to a particular technological environment or field of use (See MPEP 2106.05(h)), does not integrate the abstract idea into a practical application at Step 2A or provide an inventive concept at Step 2B.
Therefore, the additional elements of the (i) “computer system,” (ii) “processor,” (iii) “computer-readable medium,” (iv) “machine learned [models],” (v) “a multilayer perceptron comprising a plurality of layers,” (vi) “retraining the machine-learned [models],” and (vii) “client device,” do not integrate the abstract idea into a practical application at Step 2A or provide an inventive concept at Step 2B. Thus, even when viewed as a whole/ordered combination (Fig. 1-2, ¶11, ¶29, ¶48, ¶73, and ¶75 shows elements in combination.), nothing in the claims adds significantly more (i.e., an inventive concept) to the abstract idea. Thus, the claim is ineligible.
Dependent Claims 2-3, 6-9, and 22-23 recite the abstract idea of:
“. . . wherein delaying release of the order comprises: generating a second estimated delivery cost based on the plurality of tuples; and determining that the second estimated delivery cost is at least a given threshold degree lower than the first estimate” (Claim 2);
“. . . during the future time window, receiving a second order for delivery of another set of items to another address of a second user; determining that the set of items and the other set of items are to be obtained at a same retailer location; determining that the address and the other address are within a threshold distance of each other; and responsive at least in part to determining that the set of items and the other set of items are to be obtained at the same retailer and determining that the address and the other address are within a threshold distance of each other, evaluating batching the other order for delivery along with the order” (Claim 3);
“. . . apportioning delivery costs between the order and the other order, the apportioning comprising computing a batching cost savings value as a sum of a cost of delivery of the order and a separate cost of delivery of the other order, less a cost of delivering the order and the other order together” (Claim 6);
“. . . wherein the apportioning further comprises computing a ratio of a cost savings for the order with an amount of additional delivery time for the order” (Claim 7);
“. . . determining whether to batch the other order for delivery along with the order, wherein the determining comprises: generating, using the . . . counterfactual cost model, an estimated cost of delivering the order and the other order as a batch; generating incremental delivery cost savings of the estimated cost of delivering the order and the other order as a batch relative to a cost total cost of delivering the order and the other order separately; generating incremental delivery time increases of delivering the order and the other order as a batch rather than separately; and identifying whether each of the following is at least a threshold value: a ratio of an incremental delivery cost savings for the order relative to an incremental delivery time increase for the order, and a ratio of an incremental delivery cost savings for the other order relative to an incremental delivery time increase for the other order” (Claim 8);
“. . . wherein the features comprise at least one of: a location of a retailer from which to obtain items of the order, a location corresponding to the address of the user, number of items in the order, number of types of items in the order, a distance travelled for the order, a time of the order, or an estimated time of arrival of the order” (Claim 9);
“. . . wherein the . . . present cost model and the . . . counterfactual cost model are [updated] independently on respective sets of features and labels” (Claim 22); and
“. . . wherein the . . . present cost model and the . . . counterfactual model comprise distinct sets of parameters” (Claim 23).
Dependent Claims 2-3, 6-9, and 22-23, have been given the full two-prong analysis including analyzing the further elements and limitations, both individually and in combination. When analyzed individually and in combination, these claims are also held to be patent ineligible under 35 U.S.C. 101. The further limitation of Claims 2-3, 6-9, and 22-23 fail to establish claims that are not directed to an abstract idea because the further limitations of (1) delaying release of an order by estimating a second cost and applying a threshold (Claim 2), (2) receiving a second order during the future time window, determining locations are within a threshold distance and items are from the same retailer, and evaluating batching of the orders (Claim 3), (3) apportioning delivery cost between orders based on certain computations (Claims 6-7), (4) determining whether to batch the orders by estimating costs and savings of batching orders and generating delivery time increases, and identifying satisfaction of a threshold for certain calculated ratios (Claim 8), (5) limiting the extracted features to certain information (Claim 9), and (6) updating the models independently (Claim 22), and (7) the models having distinct parameters (Claim 23), are a part of the abstract idea. The elements of Claims 2-3, 6-9, and 22-23 (i.e. (i) “computer system,” (ii) “processor,” (iii) “computer-readable medium,” (iv) “machine learning,” (v) “a multilayer perceptron comprising a plurality of layers,” (vi) “retraining the machine-learned [models],” and (vii) “client device.”) fails to establish claims that are not directed to an abstract idea because the elements merely recite generic computer components similar to the generic computer components of Claim 1 and generally link the abstract idea to a particular technology or field of use (i.e. online computer environment) just as in Claim 1. The organization of the further limitations of Claims 2-3, 6-9, and 22-23 fail to integrate an abstract idea into a practical application just as discussed above for Claim 1. Additionally, performing the abstract idea of Claim 1 as recited in each of the further limitations of Claims 2-3, 6-9, and 22-23, individually or in combination, does not (1) impose any meaningful limits on practicing the abstract ideas, or (2) provide improvements to the functioning of computing systems or to another technology or technical field, just as discussed above regarding Claim 1. Therefore, Claims 2-3, 6-9, and 22-23 amount to mere instructions to implement the abstract idea (1) using generic computer components—using the computer, in its ordinary capacity, as a tool to perform the abstract idea, and (2) generally linked to a particular technology or field of use. Because the claims merely use a computer, in its ordinary capacity in a particular field of use, as a tool to perform the abstract idea cannot provide an inventive concept, the elements and limitations of Claims 2-3, 6-9, and 22-23 fail to establish that the claims provide an inventive concept, just as in Claim 1. Therefore, Claims 2-3, 6-9, and 22-23 fails the Subject Matter Eligibility Test and are consequently rejected under 35 U.S.C. 101.
Claims 10-12 and 15-18 recite elements and limitations that are substantially similar to Claims 1-3 and 6-9. Claims 1-3 and 6-9 recite a method embodied by the elements and limitations of Claims 10-12 and 15-18. Therefore, Claims 10-12 and 15-18 are rejected under 35 U.S.C. 101 just as Claims 1-3 and 6-9 are rejected under 35 U.S.C. 101 as discussed above.
Dependent Claim 14 recites the abstract idea of “. . . [updating] the counterfactual cost model based on features obtained from data log about subsequent delivery of the batch.”
Dependent Claim 14, have been given the full two-prong analysis including analyzing the further elements and limitations, both individually and in combination. When analyzed individually and in combination, these claims are also held to be patent ineligible under 35 U.S.C. 101. The further limitation of Claim 14 fail to establish claims that are not directed to an abstract idea because the further limitations of updating the model based on certain information is a part of the abstract idea. The elements of Claim 14 (i.e. (ii) “processor,” (iii) “computer-readable medium,” (iv) “machine learning,” (v) “a multilayer perceptron comprising a plurality of layers,” (vi) “retraining the machine-learned [models],” and (vii) “client device.”) fails to establish claims that are not directed to an abstract idea because the elements merely recite generic computer components similar to the generic computer components of Claim 10 and generally link the abstract idea to a particular technology or field of use (i.e. online computer environment) just as in Claim 10. The organization of the further limitations of Claim 14 fails to integrate an abstract idea into a practical application just as discussed above for Claim 10. Additionally, performing the abstract idea of Claim 10 as recited in each of the further limitations of Claim 14, individually or in combination, does not (1) impose any meaningful limits on practicing the abstract ideas, or (2) provide improvements to the functioning of computing systems or to another technology or technical field, just as discussed above regarding Claim 10. Therefore, Claim 14 amounts to mere instructions to implement the abstract idea (1) using generic computer components—using the computer, in its ordinary capacity, as a tool to perform the abstract idea, and (2) generally linked to a particular technology or field of use. Because the claims merely use a computer, in its ordinary capacity in a particular field of use, as a tool to perform the abstract idea cannot provide an inventive concept, the elements and limitations of Claim 14 fails to establish that the claims provide an inventive concept, just as in Claim 10. Therefore, Claim 14 fails the Subject Matter Eligibility Test and is consequently rejected under 35 U.S.C. 101.
Step 2A, Prong One
Independent Claim 21 recites the abstract idea of:
“. . .
a first set of parameters for a first [model],
wherein the first set of parameters includes weights and biases in . . . the first . . . model to transform an input set of features characterizing a first set of items to output an unbatched estimate on obtainment of the set of items without batching;
a second set of parameters for a second [model],
wherein the second set of parameters weights and biases in . . . the second . . . model to transform the input set of features characterizing the first set of items to output a plurality of tuples for candidate batching options,
wherein the second [model] is configured to output each tuple comprising: a candidate batch size selected from a plurality of candidate batch sizes, a batching estimate for batching the set of items with other sets at the candidate batch size in a future time window, and a likelihood associated with the candidate batch size; and
. . .
input the input set of features into the first . . . model to compute the unbatched estimate, input the input set of features into the second . . . model to compute the plurality of tuples,
determine, based at least in part on the unbatched estimate and the plurality of tuples, whether to delay release of the set of items to an unclaimed pool to attempt batching with other sets, responsive to determining to delay release, delay release of the set to the unclaimed pool for a period of time, and
responsive to receiving a second set of items during the period of time, batch the second set with the first set and
release the batch comprising the first set and the second set to the unclaimed pool,
receive a request to claim the batch comprising the first set and the second set,
transmit information on the batch comprising the first set and the second set to . . . a picker for fulfillment,
following fulfillment of the batch by the picker, compute a batched delivery cost based on a cost associated with fulfillment of the batch and a time to claim the batch from the unclaimed pool, and
[update] the second . . . model based on the batched delivery cost.”
The limitations stated above are processes/ functions that under broadest reasonable interpretation covers (1) a first set of parameters for a first model including weights and biases that are used to output an unbatched estimate based on an input, (2) a second set of parameters for a second model including weights and biases that are used to output batching tuples comprising certain data including batch size, estimated cost if the order were released, and probability associated with the estimated cost, (3) determining whether to delay the release of the order based on the estimate and the tuple, (4) delaying the release responsive to determining that the release should be delayed, (5) receiving a second order during the future time window and batching the orders, (6) releasing the batch, (7) receiving a request to claim the batch, (8) transmitting information on the batch to a picker, (9) after fulfilment, computing a cost based on certain information, and (10) updated the model based on the computed cost, all of which are mathematical calculations (i.e. using a model comprising parameters of weight and biases values to calculate outputs, computing actual cost, and updating the parameters of the model), which are mathematical concepts, an abstract idea, under MPEP 2106.04(a)(2)I, and managing personal behavior by following rules and interacting between people (i.e. determining when to release orders to pickers are at least “following rules or instructions” and communication of information is at least a “social activity”) and commercial or legal interactions (i.e. batching certain orders based on order features, predicting business metrics based on batching, releasing a batch, receiving a request, communicating necessary information for fulfilment, and providing cost feedback after fulfilment are at least “marketing or sales activities or behaviors” or “business relations”), which are certain methods of organizing human activity, an abstract idea, under MPEP 2106.04(a)(2)II. The mere the recitation of generic computer components (i.e., the “non-transitory computer readable storage medium,” “machine-learning model architected as a multilayer perception and comprised of a . . . plurality of interconnected layers,” “processor,” “retrain[ing] the machine-learned [models],” and “client device”) implementing the identified abstract idea does not take the claim out of the mathematical concepts or certain methods of organizing human activity groupings. MPEP 2106.04(d). If a claim limitation, under its broadest reasonable interpretation, covers “mathematical calculations,” “managing personal behavior or relationships or interactions between people,” and “commercial or legal interactions” but for the recitation of generic computer components, then it falls in the mathematical concepts or certain methods of organizing human activity groupings of abstract ideas. MPEP 2106.04. Therefore, Claim 1 recites an abstract idea.
Step 2A, Prong Two
The judicial exception is not integrated into a practical application. Claim 21 as a whole amounts to: (i) merely invoking generic components as a tool to perform the abstract idea or “apply it” (or an equivalent) and (ii) generally links the use of a judicial exception to a particular technological environment or field of use. The claim recites the additional elements of:
(i) “non-transitory computer readable storage medium,”
(ii) “machine-learning model architected as a multilayer perception and comprised of a . . . plurality of interconnected layers,” and
(iii) “processor,”
(iv) “retrain the second machine-learning model,” and
(v) “client device.”
The additional elements of (i) “non-transitory computer readable storage medium” (¶73 shows “a computer-readable medium comprises one or more computer-readable media that, individually or together, comprise instructions that, when executed by one or more processors, cause the one or more processors to perform, individually or together, the steps of the instructions stored on the one or more computer-readable media.” See also ¶¶74-75 discussing storing parameters and outputs.), (ii) “machine-learning model architected as a multilayer perception and comprised of a . . . plurality of interconnected layers” (Fig. 2 and ¶48 shows “The machine-learning training module 230 trains machine-learning models used by the online concierge system 140. . . Example machine-learning models include regression models, support vector machines, naïve bayes, decision trees, k nearest neighbors, random forest, boosting algorithms, k-means, and hierarchical clustering. The machine-learning models may also include neural networks, such as perceptrons, multilayer perceptrons, convolutional neural networks, recurrent neural networks, sequence-to-sequence models, generative adversarial networks, or transformers. A machine-learning model may include components relating to these different general categories of model, which may be sequenced, layered, or otherwise combined in various configurations.” See also ¶75 further discussing machine learning model. Although the specification does not explicitly disclose that a “multilayer perceptron” includes a plurality of interconnected layers, a person of ordinary skill in the art would understand that a “multilayer perceptron” is defined as having a plurality of interconnected layers.), (iii) “processor” (¶73 shows “a processor comprises one or more processors or processing units that, individually or together, perform the steps of instructions stored on a computer-readable medium”), (iv) “retrain the second machine-learned model” (¶51 shows “The machine-learning training module 230 may apply an iterative process to train a machine-learning model whereby the machine-learning training module 230 updates parameter values of the machine-learning model based on each of the set of training examples. . . . The machine-learning training module 230 updates the set of parameters for the machine-learning model based on the score generated by the loss function. For example, the machine-learning training module 230 may apply gradient descent to update the set of parameters.” See also ¶¶48-51 further showing that the claimed training is does not require a specific method of training, but could be performed by any known method for training a machine learning model.), and (v) “client device” (Fig. 1 and ¶18 shows “The picker client device 110 can be a personal or mobile computing device, such as a smartphone, a tablet, a laptop computer, or desktop computer.”), are recited at a high-level of generality, such that, when viewed as whole/ordered combination (Fig. 1-2, ¶11, ¶29, ¶48, ¶73, and ¶75 shows elements in combination.), they amount to no more than mere instruction to apply the judicial exception using generic computer components or “apply it” (See MPEP 2106.05(f)).
The (i) “non-transitory computer readable storage medium,” (ii) “machine-learning model architected as a multilayer perception and comprised of a . . . plurality of interconnected layers,” (iii) “processor,” (iv) “retrain the second machine-learned model,” and (v) “client device,” when viewed as whole/ordered combination (Fig. 1-2, ¶11, ¶29, ¶48, ¶73, and ¶75 shows elements in combination.), does no more than generally link the use of the judicial exception to a particular technological environment or field of use (i.e. online computer environment) (See MPEP 2106.05(h)).
Accordingly, these additional elements, when viewed as a whole/ordered combination (Fig. 1-2, ¶11, ¶29, ¶48, ¶73, and ¶75 shows elements in combination.), do not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea. Thus, the claim is directed to an abstract idea.
Step 2B
As discussed above with respect to Step 2A Prong Two, the additional elements amount to no more than: (i) “apply it” (or an equivalent) and (ii) generally link the use of a judicial exception to a particular technological environment or field of use, and are not a practical application of the abstract idea. The same analysis applies here in Step 2B, i.e., (i) merely invoking the generic components as a tool to perform the abstract idea or “apply it” (See MPEP 2106.05(f)) and (ii) generally linking the use of a judicial exception to a particular technological environment or field of use (See MPEP 2106.05(h)), does not integrate the abstract idea into a practical application at Step 2A or provide an inventive concept at Step 2B.
Therefore, the additional elements of the (i) “non-transitory computer readable storage medium,” (ii) “machine-learning model architected as a multilayer perception and comprised of a . . . plurality of interconnected layers,” (iii) “processor,” (iii) “processor,” (iv) “retrain the second machine-learned model,” and (v) “client device,” do not integrate the abstract idea into a practical application at Step 2A or provide an inventive concept at Step 2B. Thus, even when viewed as a whole/ordered combination (Fig. 1-2, ¶11, ¶29, ¶48, ¶73, and ¶75 shows elements in combination.), nothing in the claims adds significantly more (i.e., an inventive concept) to the abstract idea. Thus, the claim is ineligible.
Reasons for No Art Rejection
Claims 1-3, 6-12, 14-18, and 21-23 are not rejected over the prior art of record. Applicant has an actual and effective filing date of 02/16/2024. Prior art, as defined by 35 U.S.C. 102 and MPEP 2153, excludes art with a common assignee or inventor published within a 1-year grace period (i.e. after 02/16/2023), from being considered “prior art.”
The Closest prior art of record is:
US-20190130350-A1 (“Nguyen” Published on 05/02/2019, Later granted as US-11436554-B2, Assigned to Uber Technologies Inc.);
US-20190132702-A1 (“Ehsani” Published on 05/02/2019, Assigned to Uber Technologies Inc.);
US-20180308038-A1 (“Zhou” Published on 08/16/2022, Later granted as US-11416792-B2, Assigned to Uber Technologies Inc.);
US-20210081880-A1 (“Bivins” Published on 03/18/2021, Later granted as US-11216770-B2, Assigned to Uber Technologies Inc.);
US-20200342517-A1 (“Rajkhowa” Published on 10/10/2023, Later granted as US-11783403-B2, Assigned to Walmart Apollo, LLC.);
US-20150227888-A1 (“Levanon” Published on 07/12/2015, Assigned to Dragontail Systems, Ltd.);
US-20190180229-A1 (“Phillips” Published on 06/13/2019, Later granted as US-10783482-B2, Assigned to Capital One Services, LLC.)
US-8756119-B1 (“Andrews” Published on 06/17/2014, Assigned to T Mobile Innovations, LLC.);
US-20220045951-A1 (“Ahn” Published on 02/10/2022, Later granted as US-11818047-B2, Assigned to Coupang Corp.);
US-20220292580-A1 (“Putrevu” Published on 09/15/2022, Later granted as US-11803894-B2, Assigned to (common assignee) Maplebear, Inc.);
US-20200410864-A1 (“Ripert-864” Published on 12/31/2020, Later granted as US-11580860-B2, Assigned to (common assignee) Maplebear, Inc.);
US-20190114583-A1 (“Ripert-583” Published on 04/18/2019, Later granted as US-10818186-B2, Assigned to (common assignee) Maplebear, Inc.);
WO-2020131987-A1 (“Prestezog” Published on 06/25/2020, Assigned to Zume, Inc.);
CN-111582612-A (“Shi”) Published on 08/25/2020, Assigned to Lazas Network Technology Shanghai Co Ltd.); and
“Multilayer Perceptron in Machine Learning: A Comprehensive Guide” (Jaiswal, 02/14/2024, https://web.archive.org/web/20240214003028/https://www.datacamp.com/tutorial/multilayer-perceptrons-in-machine-learning).
The Following is an examiner’s statement of reasons for not applying prior art:
Nguyen Fig. 1 shows a method of matching orders with drivers (i.e. pickers) that utilizes batching of orders together. Fig. 1 and ¶44 shows a “requestor status store 132” which stores the status of the order. The “request handling component 128” times when an order should be released to “Matching Component 140,” which matches the orders to the drivers (i.e. making unclaimed orders available to pickers). ¶35. The “request handling component 128” works with the “batch decision logic 138” that applies certain rules to batch orders together based in part on supply and demand. ¶39. The rules could include the orders being from the same supplier, a timing component that ensures it would be practicable to combine the orders. ¶¶39-43.
Based on the “requester status store 132” and “historical information 159,” the “model development component 158” creates a plurality of “models 153.” ¶¶49-56. The models use “requester status store 132” as inputs and forecasts several parameters including future order demand over a certain time frame. ¶¶49-56. The forecasts includes “timing parameter 167” is used by “request handling component 128,” which can be “relaxed” to provide additional time for batching orders. ¶¶71-72. However, Nguyen does not discuss a specific algorithm for how the “timing parameter 167” is relaxed.
The “models 153” also output a “service value 171” (i.e. cost for order delivery) indicating the charge for delivery. ¶63. However, ¶63, ¶66, and ¶75 shows that the “service value 171” can be changed based on supply and demand, or distance traveled to complete delivery. Although batching would affect supply and demand, or distance traveled, Nguyen does not disclose counterfactual simulations or probabilities associated with the cost. Thus, although the models of Nguyen may be capable of such outputs, Nguyen does not provide the specific level of detail disclosed in Applicant’s claims.
Ehsani discloses a system and method similar to that of Nguyen.
Zhou shows grouping item deliver orders based on timing such that pending orders can be grouped with new orders. Fig. 1 and ¶¶36-37. The grouped orders “127” are delivered via “service provider 192” who receives a route provided by “provider routing and selection engine 120.” ¶37. The fees for delivery are associated with each “entity 116” (i.e. item provider). ¶35.
Bivins includes a “dynamic cost optimizer 156” that can delay the release of an order based on the expected cost and a threshold cost by forecasting supply and demand with “forecasting engine 160.” Fig. 1, ¶¶31-32, and ¶43. However, this delay relates to delaying the release of the order to the supplier (e.g. the restaurant). Additionally, Bivins does not mention batching multiple orders together.
Rajkhowa shows an operation to combine certain orders for a single deliver driver, or to have the deliver driver transport a single order. ¶29. Of the parameter’s considered, ¶¶29-31 shows that a time window constraint is used to ensure that combining orders ensures delivery within the scheduled timeslot (see ¶¶48-53 discussing scheduling the time slot). Additionally, the schedule considers the capacity of the deliver vehicle and limits the number of orders that may be batched ¶¶54-55.
Although batch pricing may be different for the delivery driver (¶33), Rajkhowa does not discuss the calculation of the price specifically. Thus, Rajkhowa utilizes certain criteria to determine if orders should be batched. However, Rajkhowa does not consider the counterfactual options of delaying the release of an order to be combined later.
Levanon ¶54 shows a “deliver delay parameter” that quantifies the extra time of optionally adding a second order to a first delivery order, batching the orders if the “deliver delay parameter” satisfies a threshold. Fig. 3A, ¶¶46-47, and ¶¶59-60 shows a method of determining optional sets of orders and then assigning the orders to a courier. The assignment process is optimized by utilizing a simulation of the delivery process. ¶¶78-82. Although Levanon considers the delay caused by combining orders, Levanon does not control this delay by optimizing cost and probability of selection.
Phillips Fig. 1A-1B and 4 shows a scheduling platform that allocates orders to couriers to be delivered from a product location to the user. Fig. 4 and ¶¶71-82 shows the determination of a delay of the order based on fulfillment time (i.e. item preparation time), estimated delivery time, and an additional delay that functions as a factor of safety. Although ¶70 considers the combination of items within an order, Phillips does not consider combination or batching of orders.
Andrews Fig. 1 shows applying “aggregation rules” to combine orders for more efficient (e.g. cost effective) shipping. Fig. 1 and C06L54-C09L30. Example rules include delaying release of an order so that it may be combined with future orders to reduce shipping cost (e.g. optimizing for full capacity shipments). C02L52-C03L05. Based on the rules, the shipments are released; however, the releasing of the orders in Andrews is the actual releasing to the courier, not merely releasing orders such that they become available to be claimed. C08L66-C09L30. Additionally, Andrews does not provide specifics as to the rules that cause delays (e.g. cost and probability analysis).
Ahn discloses a system and method for pooling deliver requests upon certain conditions. Fig. 3 and ¶22. Incoming request can be delayed if the pool is already at capacity. ¶¶3-4. However, this delay is more related to data transmission (Fig. 4 and ¶¶99-101), to address network congestion.
Putrevu uses a machine learning model to predict probability of fulfillment of an order (e.g. probability that a certain warehouse has the item). To increase efficiency, “flexible fulfillment” orders can be grouped based on a machine learned cost model. ¶¶33-34. The grouping applies a specified “time window” constraint for “flexible” orders which ensures delivery by a certain time. ¶33. See also Fig. 5 and ¶¶49-62. Although similar to the instant claims, much of the specifics are not taught by Petrevu including that the probability of Putrevu (1) is not dependent upon deliver cost or batching of orders and (2) is not associated with the estimated delivery cost.
Ripert-864 shows batching orders based on supplier location and service modeling, then based on the batches, select a delivery driver. Fig. 4 and ¶35. The models include estimating times and number of bags for orders. ¶36. The orders can be re-allocated to deliver drivers based on updated information. Fig. 5 and ¶¶38-39. Ripert-864 does not consider delaying release of an order to enable batching, but instead executes the batching procedure for each order.
Ripert-583 discloses a system and method similar to that of Ripert-864.
Prestezog shows grouping delivery orders to schedule a deliver plan. ¶¶30-31. However, the delays are primarily for delaying preparation of the item. ¶31 and ¶69. Fig. 6-7 and ¶¶72-79 shows combining orders based on scores or deliver cost. However, the orders are merely assigned to drivers (Fig. 5A-5C and ¶¶85-90) and probability of batching is not explicitly considered. batching orders.
Shi shows scheduling deliveries by determining a “target scheduling policy combination” and allocating the orders to delivery drivers. Fig. 1 and Page 8. The orders are allocated based on delivery costs from candidate deliver resources (i.e. delivery drivers) based on the lowest cost deliver resource. Fig. 4 and Page 13-14. batching orders. However, Shi does not teach the specific parameters considered in the instant claims.
Jaiswal shows that each neuron of the hidden layers in a multilayer perceptron is a mathematical equation that outputs a weighted sum of the inputs, with an additional “bias” factor.
Generally, the closest prior art teaches either (1) grouping orders for delivery (Nguyen, Ehsani, Zhou, Rajkhowa, Levanon, Andrews, Ripert-864, Ripert-583, Prestezog, and Shi), or (2) providing modeling of order deliveries (Nguyen, Ehsani, Bivins, Rajkhowa, Phillips, Ahn, Putrevu, and Shi), and (3) multilayer perceptron neural networks (Jaiswal), without the specific details claimed.
With respect to representative independent Claim 1, the closest prior art, taken individually and in an ordered combination, does not explicitly or implicitly disclose the specific ordered combination of elements that include “providing the set of features as input to a machine-learned present cost model to obtain a first estimate representing an estimated delivery cost if the order were released from the batching candidate pool to the unclaimed order pool without attempting batching . . . ; providing the set of features as input to a machine-learned counterfactual cost model to obtain a plurality of tuples, . . . each tuple comprising: a possible batch size, an estimated delivery cost if the order were released to the unclaimed order pool during a given future time window allowing for a possibility of batching the order with orders of other users, and a probability associated with the estimated delivery cost; determining, based at least in part on the first estimate and on the tuples, whether to delay release of the order to the unclaimed order pool in order to attempt batching of the order.”
Independent Claims 10 and 21 recite limitations substantially similar to the novel and non-obvious limitations of representative Claim 1. Thus, independent Claims 10 and 21 are not rejected based on the prior art.
Dependent Claims 2-3, 6-9, 11-12, 14-18, and 22-23 depend on Claims 1 and 10, and therefore are also not rejected under 35 U.S.C. 102 or 35 U.S.C. 103 via dependency.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to MATTHEW PARKER GOODMAN whose telephone number is (571) 272-5698. The examiner can normally be reached on Monday-Thursday from 9:30 AM ET to 6:00 PM ET. If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Jeffrey Zimmerman, can be reached at telephone number (571) 272-4602. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of an application may be obtained from the Patent Application Information Retrieval (PAIR) system. Status information for published applications may be obtained from either Private PAIR or Public PAIR. Status information for unpublished applications is available through Private PAIR only. For more information about the PAIR system, see http://portal.uspto.gov/external/portal. Should you have questions about access to the Private PAIR
system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free).
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.
/MATTHEW PARKER GOODMAN/Examiner, Art Unit 3628
/JESSICA LEMIEUX/Supervisory Patent Examiner, Art Unit 3626
Prosecution Insights