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 .
Response to Arguments
The objections to the Drawings and Specification from the Non-Final Office Action filed on 9/10/2025 are maintained as they do not appear to be addressed by the Applicant.
The objection to claim 30 is withdrawn based on amendment filed on 1/12/2026.
The 35 U.S.C. 112(b) rejection of claims 1, 30 and 32 for claim term “relevant” does not appear to be addressed. Therefore, the 35 U.S.C. 112(b) rejections associated with this term is maintained.
The 35 U.S.C. 112(b) rejection of claim 11 for claim terms “perception router” does not appear to be addressed. Therefore, the 35 U.S.C. 112(b) rejection associated with this term maintained.
The 35 U.S.C. 112(b) rejection with respect to claims 1, 30 and 32 for previous claim term “respective fraction” has been withdrawn based on the amendment filed on 1/12/2026.
Applicant’s arguments in light of the amendment filed on 1/12/2026, with respect to 35 U.S.C. 112(a) rejections have been fully considered and are persuasive. The lack of written description and lack of enablement rejections of claims 1-16, 19-23, 25-28, 30 and 32 have been withdrawn.
Applicant’s arguments in light of the amendment filed on 1/12/2026 with respect to the 35 U.S.C. 102 and 35 U.S.C. 103 rejections of claims 1-16, 19-23, 25-28, 30 and 32 have been fully considered and are persuasive. Therefore, the prior rejections have been withdrawn. However, upon further consideration, a new ground(s) of rejection is made in view of an updated prior art search on the amended claim filed on 1/12/2026.
Drawings
The drawings are objected to because:
In fig. 1, “Corrdinator” should be ‘Coordinator’
In fig. 1, the numerals specified in ¶41 are not shown
Corrected drawing sheets in compliance with 37 CFR 1.121(d) are required in reply to the Office action to avoid abandonment of the application. Any amended replacement drawing sheet should include all of the figures appearing on the immediate prior version of the sheet, even if only one figure is being amended. The figure or figure number of an amended drawing should not be labeled as “amended.” If a drawing figure is to be canceled, the appropriate figure must be removed from the replacement sheet, and where necessary, the remaining figures must be renumbered and appropriate changes made to the brief description of the several views of the drawings for consistency. Additional replacement sheets may be necessary to show the renumbering of the remaining figures. Each drawing sheet submitted after the filing date of an application must be labeled in the top margin as either “Replacement Sheet” or “New Sheet” pursuant to 37 CFR 1.121(d). If the changes are not accepted by the examiner, the applicant will be notified and informed of any required corrective action in the next Office action. The objection to the drawings will not be held in abeyance.
Specification
The disclosure is objected to because of the following informalities:
In ¶80, “Each narrow AI agent is narrow in the senses that…” should be ‘Each narrow AI agent is narrow in the sense that…’.
In ¶88, “the Narrow AI agent may executed step 130” should be ‘the Narrow AI agent may execute step 130’.
In ¶111, “…or in any suitable sub combination” should be ‘…or in any suitable sub-combination’.
Appropriate correction is required.
Claim Rejections - 35 USC § 112(b) or 35 USC § 112, second paragraph
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 1-16, 19-23, 25-28, 30 and 32 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Per claims, 1, 30 and 32, the term “relevant” in the claims is a relative term which renders the claim indefinite. The term “relevant” is not defined by the claim, the specification does not provide a standard for ascertaining the requisite degree, and one of ordinary skill in the art would not be reasonably apprised of the scope of the invention. The term is a term of degree and is subjective.
Per claims, 1, 30 and 32, the term “respective fraction” in the claims is a relative and unbounded term which renders the claim indefinite. The term “respective fraction” is not defined by the claim, the specification does not provide a standard for ascertaining the requisite degree, and one of ordinary skill in the art would not be reasonably apprised of the scope of the invention. It can mean an infinitesimally small portion or a substantial portion, e.g., 51%.
Per claim 11, the limitation "the perception router" has insufficient antecedent basis for this limitation. To expedite prosecution, Examiner interprets this to refer to the perception unit.
The dependent claims are rejected based on dependency to the rejected independent claims.
Claim Rejections - 35 USC § 103
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.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claims 1-3, 6-16, 19-23, 25-28, 30 and 32 are rejected under 35 USC 103 as being unpatentable over US Pat. Pub. No. 2020/0249637 to Wee et al. (hereinafter Wee) in view of AIBA: An AI Model for Behavior Arbitration in Autonomous Driving to Trasnea et al. (hereinafter Trasnea).
Per claim 1, Wee discloses A method for operating an ensemble of narrow AI agents (Abstract and fig. 6:82…ensemble control system which combines different types of plant control using a “plurality of subcontrollers”, each construed as a narrow AI agent, and each of which “outputs action for the plant control based on a prediction result by a predictor", the ensemble control system having “a combiner or switch 82 combines or switches actions to maximize prediction or control performance…"; fig. 5:120 …three types of subcontrollers including learned subcontroller, model predictive subcontroller, and alternative subcontroller), the method comprises:
obtaining one or more sensed information units ([0030]…"The outputs or observations 110 can be, for example, the state of the plant 106 or variables related to the environment which are acquired by the sensor"; [0050]…”predictors 101 and subcontrollers receive state measurements such as position, speed and other observations from the plant 106 (e.g. vehicle)”);
processing the one or more sensed information units, by the one or more relevant narrow AI agents, to provide one or more narrow AI agent outputs (fig. 2 and [0030]…each subcontroller 102-104 is associated with a predictor 111-113 that process plant outputs/observations, "The predictors 101 are associated to each subcontroller, and given outputs or observations 110 from the plant 106, the predictors 101 calculate predictions which are sent to the subcontrollers"; fig. 3:S101-S103 and [0032]…”In the case of autonomous driving for example, each of the subcontrollers will output their calculated "best" steering angle and acceleration, e.g., (0.785 rad, 2.5 m/sec^2)”);
processing, by an intermediate result unit, the one or more narrow AI agent outputs to provide an intermediate result; wherein the processing comprises applying a function selected out of selecting a narrow AI agent output, averaging the narrow AI agent outputs or performing a weighted sum of the narrow AI agent outputs (fig. 2:105 and [0046]…the classifier/combiner gets outputs/intermediate results from the subcontrollers, "has the capability to merge the different control inputs (e.g., by averaging as mentioned above), and regarding just choosing the control actions between the different subcontrollers (e.g., by voting or using confidence levels)"; [0044]…"the classifier/combiner 105 may also decide the best control operation from the outputs of the subcontrollers by voting, if for categorical actions, or by averaging, for numerical actions"; [0043]…"The classifier/combiner 105 may decide the best control operations to actuate by…choosing the action that minimizes a weighted sum…"); and
generating a response, by a response unit, based on the intermediate result (fig. 5…the final control action 107 sent out to the plant 106 for actuation is sent by main controller which generates the response based on classifier/combiner 105; fig. 5:105 and [0072]…the combiner produces the intermediate result to be sent to the main controller, "the combiner or switch 82 may compute a best control action to be actuated from the set of the control actions and the predicted actions output by the different subcontrollers 81";).
While not explicitly stated, Wee intrinsically discloses that the subcontrollers may be selectively activated based on the task at hand by stating that, "at least one subcontroller is required to be active at each time and the others may be inactive. Which subcontrollers are active or inactive depends on the task and the resulting choice of the ensemble method" ([0040]), and further teaches that the classifier/combiner "may keep the historical performance of each subcontroller in different kinds of control scenarios (such as driving maneuvers)" ([0045]). This establishes scenario-dependent selection of subcontrollers. Additionally, Wee's autonomous driving application outputs driving instructions in the form of "the front wheel steering angle and longitudinal speed" (fig. 3 and [0049]).
Wee does not expressly disclose, but with Trasnea does teach:
determining, by a perception unit and based on the one or more sensed information units, one or more scenarios (Trasnea: Section 3…the perception system processes sensor data to "transform the scene into a collection of objects" including "traffic participants, pedestrians and buildings" classified as "static objects (lanes, traffic signs, buildings, etc.) and dynamic objects (cars, pedestrians, etc.)"; Section 4…the perception module uses occupancy grids for "environment perception and navigation" to identify the driving context and scenario, “We used such a representation in the previous work for driving context classification”)
determining one or more relevant narrow AI agents of the ensemble, that are trained to output driving instructions at the presence of the one or more scenarios; wherein the ensemble is relevant to a first plurality of scenarios (Trasnea: Section 3, Equation 9…the system selects behaviors based on detected scenarios where the argmax function determines the optimal behavior to activate; Section 4, Table 1…multiple specialized driving behaviors including LaneFollow, LaneChangeRight, LaneChangeLeft, and Stop are each trained for specific driving scenarios and output driving commands such as steering and speed adjustments);
wherein each narrow AI agent is relevant to less than one percent of the first plurality of scenarios (Trasnea: Section 4…each behavior is specialized for a specific driving scenario type; Wee: [0036]…"there can be many learned subcontrollers in the system, each of which might have been learned using different machine learning techniques, may be based on different predictive models, or trained using different datasets"; the autonomous driving domain intrinsically involves a vast number of scenarios as combinations of road types, weather conditions, traffic parameters, obstacle types, and contextual factors creating hundreds to thousands of distinct scenarios; scaling the ensemble so that each narrow agent specializes in less than one percent of the total scenario space is a predictable engineering design choice to achieve higher performance through increased specialization).
Wee and Trasnea are analogous art because they are both within the same field of endeavor, specifically artificial intelligence systems for autonomous vehicle control and decision-making. They address the same problem solving area of combining multiple specialized AI models or behaviors to generate driving commands for autonomous vehicles. Wee describes an ensemble of subcontrollers for autonomous driving (Wee: FIG. 3…autonomous driving application with steering angle and longitudinal speed as controlled variables), and Trasnea provides a behavior arbitration model for autonomous driving that selects from multiple specialized driving behaviors based on perceived scenarios (Trasnea: Section 1…"arbitrate between different driving strategies, or vehicle behaviors, based on understanding of the relations between the scene objects").
Before the effective filing date of the claimed invention, it would have been obvious to a person of ordinary skill in the art to modify the ensemble control system of Wee by incorporating the perception-based scenario detection and behavior arbitration mechanism of Trasnea, such that the ensemble of subcontrollers in Wee would include a perception unit that determines driving scenarios from sensor data and selects relevant specialized subcontrollers trained for those specific scenarios.
The suggestion/motivation for doing so would have been that Wee explicitly teaches scenario-dependent subcontroller activation ([0040]…"Which subcontrollers are active or inactive depends on the task…") but does not specify how the task or scenario is determined from sensor data. Trasnea provides exactly this mechanism — a perception system that analyzes sensor data to identify driving scenarios and then selects appropriate driving behaviors (Trasnea: Section 3…the perception system identifies objects and threats to determine the current driving scenario). Combining these teachings would yield the predictable result of an ensemble control system that autonomously identifies the current driving scenario from sensor data and activates the appropriate specialized subcontrollers, thereby improving the system's ability to respond optimally to diverse driving situations. Furthermore, regarding the less than one percent limitation, a PHOSITA would have found it obvious to scale the number of specialized agents to handle the combinatorial complexity of real-world driving scenarios (road type, weather, traffic, obstacles, time of day, pedestrian behavior, etc.), where each highly specialized agent handles only a small fraction of the total scenario space, as this approach follows directly from the principle of increased specialization taught by both Wee and Trasnea.
Per claim 2, Wee combined with Trasnea discloses claim 1. Wee combined with Trasnea further teaches for at least some of the narrow AI agents the respective fraction is smaller than one percent of the first plurality of scenarios (Wee: [0036]…many learned subcontrollers each trained on different datasets, “Note that there can be many learned subcontrollers 102 in the system, each of which might have been learned using different machine learning techniques, may be based on different predictive models, or trained using different datasets.”; Trasnea: Section 4…each specialized behavior handles only specific scenario types; the same rationale regarding scaling the ensemble applies as discussed in claim 1). The rationale to combine Trasnea with Wee is the same as the parent claim.
Per claim 3, Wee combined with Trasnea discloses claim 1. Wee combined with Trasnea further teaches wherein a number of narrow AI agents relevant to one of the first plurality of scenarios differs from a number of narrow AI agents relevant to another of the first plurality of scenarios (Wee: [0040]…"it is possible to have more than one subcontroller for each type of subcontroller (e.g., two active learned subcontrollers 102 and two active model predictive subcontrollers 103)"; the number of active subcontrollers varies depending on the task; Trasnea: Section 3, Equation 9…different scenarios activate different numbers of relevant behaviors based on threat assessment). The rationale to combine Trasnea with Wee is the same as the parent claim.
Per claim 6, Wee combined with Trasnea discloses claim 1. Wee combined with Trasnea further teaches wherein each narrow AI agent is trained to respond to a respective fraction of the first plurality of scenarios (Wee: [0036]…"there can be many learned subcontrollers 102 in the system, each of which might have been learned using different machine learning techniques, may be based on different predictive models, or trained using different datasets"; Trasnea: Section 4…each behavior is trained for specific scenario types). The rationale to combine Trasnea with Wee is the same as the parent claim.
Per claim 7, Wee combined with Trasnea discloses claim 1. Wee combined with Trasnea further teaches wherein at least some of the narrow AI agents comprise at least a portion of a neural network (Wee: [0034]…learned subcontrollers are trained using "deep reinforcement learning or other machine learning models"; [0031]…predictors "may employ any machine learning technique such as kernel methods or deep neural networks").
Per claim 8, Wee combined with Trasnea discloses claim 1. Wee combined with Trasnea further teaches wherein the determining of the one or more relevant narrow AI agents comprises determining one or more obtained scenarios that are related to the one or more sensed information units, and determining a relevancy of the narrow AI agents based on a relationship between the one or more obtained scenarios and an association between the first plurality of scenarios and the narrow AI agents (Trasnea: Section 3…the perception system identifies objects and their properties from sensor data to determine the current scenario; the threat simulation function evaluates each behavior against the identified scenario to determine relevancy; Wee: [0045]…the classifier/combiner maintains "the historical performance of each subcontroller in different kinds of control scenarios" establishing associations between scenarios and appropriate subcontrollers). The rationale to combine Trasnea with Wee is the same as the parent claim.
Per claim 9, Wee combined with Trasnea discloses claim 8. Wee combined with Trasnea further teaches wherein the determining of the one or more relevant narrow AI agents comprises determining that a narrow AI agent is relevant when the narrow AI agent is associated to any of the one or more obtained scenarios (Trasnea: Section 3, Equation 9…a behavior is selected as relevant when it addresses the detected scenario; the system evaluates which behaviors are applicable to the currently obtained scenario). The rationale to combine Trasnea with Wee is the same as the parent claim.
Per claim 10, Wee combined with Trasnea discloses claim 8. Wee combined with Trasnea further teaches wherein the association between the first plurality of scenarios and the narrow AI agents is manually determined (Trasnea: Section 3…the driving behaviors (LaneFollow, LaneChangeRight, LaneChangeLeft, Stop) are predefined and manually associated with specific threat types and driving scenarios; the behavior set is designed by human engineers based on driving domain knowledge). The rationale to combine Trasnea with Wee is the same as the parent claim.
Per claim 11, Wee combined with Trasnea discloses claim 8. Wee combined with Trasnea further teaches wherein the association between the first plurality of scenarios and the narrow AI agents is determined based on previous determining made by the perception router (Wee: [0045]…the classifier/combiner "may keep the historical performance of each subcontroller in different kinds of control scenarios" to establish "confidence levels regarding the use of input actions from specific subcontrollers"; the association between scenarios and agents is refined based on previous observations by the system).
Per claim 12, Wee combined with Trasnea discloses claim 1. Wee combined with Trasnea further teaches wherein the determining of the one or more relevant narrow AI agents comprises determining one or more obtained scenario parts that are related to the one or more sensed information units, and determining a relevancy of the narrow AI agents based on a relationship between the one or more obtained scenario parts and an association between the first plurality of scenarios and the narrow AI agents (Trasnea: Section 3…the perception system decomposes the scene into individual objects including traffic participants, pedestrians, and static objects; the threat simulation function evaluates individual object threats as scenario parts to determine which behaviors are relevant). The rationale to combine Trasnea with Wee is the same as the parent claim.
Per claim 13, Wee combined with Trasnea discloses claim 12. Wee combined with Trasnea further teaches wherein at least some of the obtained scenario parts are associated with one or more objects that were sensed in the one or more sensed information units (Trasnea: Section 3…the perception system identifies objects including "traffic participants, pedestrians and buildings" that are directly sensed; each object constitutes a scenario part associated with sensed data). The rationale to combine Trasnea with Wee is the same as the parent claim.
Per claim 14, Wee combined with Trasnea discloses claim 1. Wee combined with Trasnea further teaches feeding the one or more sensed information units to each one of the one or more relevant narrow AI agents (Wee: [0030]…"given outputs or observations 110 from the plant 106, the predictors 101 calculate predictions which are sent to the subcontrollers", the observations are provided to each active subcontroller through its associated predictor).
Per claim 15, Wee combined with Trasnea discloses claim 1. Wee combined with Trasnea further teaches determining which part of the one or more sensed information units to send to each relevant narrow AI agent (Wee: [0031]…"each predictor 101 computes state predictions that are required by each type of subcontroller"; different predictors process different aspects of the observations relevant to their associated subcontroller; Trasnea: Section 3…the perception system identifies specific objects and their properties relevant to each behavior). The rationale to combine Trasnea with Wee is the same as the parent claim.
Per claim 16, Wee combined with Trasnea discloses claim 1. Wee combined with Trasnea further teaches wherein a narrow AI agent output is a command (Wee: FIG. 3…each subcontroller outputs a control command such as steering angle and acceleration; the output is a direct control command for the vehicle).
Per claim 19, Wee combined with Trasnea discloses claim 1. Wee combined with Trasnea further teaches wherein a narrow AI agent output is a suggested response of the response unit (Wee: [0043]…the subcontroller outputs are candidate actions that are evaluated by the classifier/combiner; each subcontroller output is effectively a suggested control action that may or may not be selected as the final response).
Per claim 20, Wee combined with Trasnea discloses claim 1. Wee combined with Trasnea further teaches intermediate result unit is configured to select at least one selected narrow AI agent output of the one or more narrow AI agent outputs (Wee: [0043]…the classifier/combiner decides "the best control operation to actuate by comparing the values of certain performance measures" and selecting the best action").
Per claim 21, Wee combined with Trasnea discloses claim 1. Wee combined with Trasnea further teaches intermediate result unit is configured to average the one or more narrow AI agent outputs (Wee: [0046]…the classifier/combiner has "the capability to merge the different control inputs (e.g., by averaging)").
Per claim 22, Wee combined with Trasnea discloses claim 1. Wee combined with Trasnea further teaches wherein each narrow AI agent output of the one or more narrow AI agent outputs is associated with a time period (Wee: [0050]-[0054], FIG. 3…the ensemble control system operates in a repeated time-stepped loop (FIG. 3, "REPEAT" after S105). At each time step: (S101) predictors and subcontrollers receive measurements from the plant, (S102-S103) each subcontroller computes control actions, (S104) the classifier/combiner selects or combines actions, and (S105) the plant actuates the result. Each subcontroller output (e.g., steering angle and acceleration per [0032]) is therefore explicitly associated with the specific time step at which it was computed; Trasnea: Section 3, Equations 5-8…the threat simulation operates over a specific time horizon H = [tc, tc + delta] where tc is the current time, producing behavior recommendations associated with that specific time window). The rationale to combine Trasnea with Wee is the same as the parent claim.
Per claim 23, Wee combined with Trasnea discloses claim 1. Wee combined with Trasnea further teaches wherein different narrow AI agent outputs of the one or more narrow AI agent outputs are associated with different time periods, wherein the intermediate result unit is configured to generate an intermediate result that is responsive, at each of the different time periods, to a narrow AI agent output related to the time period (Wee: [0040]…"at least one subcontroller is required to be active at each time and the others may be inactive. Which subcontrollers are active or inactive depends on the task"; [0045]…the classifier/combiner "may keep the historical performance of each subcontroller in different kinds of control scenarios" and adjusts which subcontrollers' outputs to use based on scenario changes over time. As the driving scenario evolves across successive time steps (e.g., transitioning from highway driving to an intersection approach), different subcontrollers become relevant and produce outputs at different time periods. The classifier/combiner is configured to generate the intermediate result at each time period by processing the outputs from the subcontrollers that are active during that specific time period, as shown in the iterative loop of FIG. 3).
Per claim 25, Wee combined with Trasnea discloses claim 22. Wee combined with Trasnea further teaches wherein the intermediate result comprises instructions for operating a robot (Wee: FIG. 3…the intermediate result is a control action comprising steering angle and speed for operating an autonomous vehicle, which is a robotic system).
Per claim 26, Wee combined with Trasnea discloses claim 1. Wee combined with Trasnea further teaches wherein the processing by the intermediate result unit comprises combining multiple narrow AI agent outputs by applying risk reduction optimization (Wee: [0043]…the classifier/combiner compares "values of certain performance measures...such as distance to surrounding objects, comfort level, safety and energy consumption" and chooses "the action that minimizes a weighted sum" of these measures, which constitutes risk reduction optimization; Trasnea: Section 3, Equation 9…the behavior arbitration selects the behavior that maximizes threat resolution, which is inherently a risk reduction optimization). The rationale to combine Trasnea with Wee is the same as the parent claim.
Per claim 27, Wee combined with Trasnea discloses claim 1. Wee combined with Trasnea further teaches wherein the determining of the one or more relevant narrow AI agents of the ensemble is based on the one or more sensed information units and based on at least one additional parameter (Wee: [0045]…the classifier/combiner uses both current observations and "the historical performance of each subcontroller in different kinds of control scenarios" as additional parameters for determining which subcontrollers to activate).
Per claim 28, Wee combined with Trasnea discloses claim 27. Wee combined with Trasnea further teaches wherein the at least one additional parameter is a purpose assigned to the method (Wee: [0040]…"Which subcontrollers are active or inactive depends on the task"; the task assignment is a purpose parameter that influences which subcontrollers are selected).
Claims 30 and 32 are substantially similar in scope and spirit as claim 1, reciting a non-transitory computer readable medium and a computerized system, respectively, with the same functional limitations. Therefore the rejections of claim 1 are applied accordingly.
Claims 4 and 5 are rejected under 35 USC 103 as being unpatentable over Wee in view of Trasnea and further in view of Ensemble Selection from Libraries of Models to Caruana et al. (hereinafter Caruana).
Per claim 4, Wee combined with Trasnea discloses claim 1.
Wee combined with Trasnea does not expressly disclose, but with Caruana does teach: wherein a number of narrow AI agents exceeds one thousand (Caruana: Abstract, Section 2…the ensemble selection method operates on "libraries of thousands of models" generated using different learning algorithms, parameter settings, and feature subsets; the method demonstrates that constructing and selecting from very large model libraries yields better ensemble performance)
Per claim 5, Wee combined with Trasnea discloses claim 1.
Wee combined with Trasnea and Caruana does not expressly disclose wherein a number of narrow AI agents exceeds one hundred thousand. However, it would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to scale the number of narrow AI agents to exceed one hundred thousand. Caruana teaches libraries of thousands of models (not appearing to set a limit), and Wee teaches that the number of subcontrollers is scalable. In the autonomous driving domain, the combinatorial space of driving scenarios (weather conditions multiplied by road types multiplied by traffic parameters multiplied by obstacle types multiplied by time of day multiplied by contextual factors) creates hundreds of thousands of distinct situations. A PHOSITA would recognize that scaling the ensemble to one hundred thousand or more specialized agents to cover this vast scenario space is a predictable design choice that follows from the teachings of Wee, Trasnea, and Caruana, yielding the predictable result of higher specialization and better driving performance across diverse scenarios.
Wee, Trasnea, and Caruana are analogous art because they are all within the field of machine learning systems that combine multiple specialized models or agents to achieve better performance than individual models. Wee teaches an ensemble of subcontrollers for autonomous driving (Wee: FIG. 3), Trasnea teaches behavior arbitration with specialized driving behaviors (Trasnea: Section 1), and Caruana teaches methods for selecting and combining models from very large model libraries (Caruana: Abstract…"a method for constructing ensembles from libraries of thousands of models").
Before the effective filing date of the claimed invention, it would have been obvious to a person of ordinary skill in the art to apply Caruana's technique of building and selecting from libraries of thousands of models to the ensemble control system of Wee as modified by Trasnea, thereby scaling the number of specialized driving agents to exceed one thousand.
The suggestion/motivation for doing so would have been that Caruana demonstrates that ensemble performance improves with larger model libraries, as the ensemble selection process can identify the best combination of models from a larger pool (Caruana: Section 2…the selection process iteratively adds models that maximize ensemble performance from the library). Wee's ensemble of subcontrollers would benefit from this scaling principle, as a larger pool of specialized driving agents allows the system to better cover the vast space of driving scenarios and select the most appropriate agents for each situation.
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.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Kamran Afshar can be reached at (571) 272-7796. 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.
/ALAN CHEN/Primary Examiner, Art Unit 2125