DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA . This action is in response to an RCE filed on March 30th, 2026. Claims 1-21 are pending in the current application.
Continued Examination Under 37 CFR 1.114
A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on March 30th, 2026 has been entered.
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 1-21 are rejected under 35 U.S.C. 101 because the claimed invention is directed
to an abstract idea without significantly more.
Regarding claim 1, under Step 1 of the Subject Matter Eligibility Test of Products and Processes, claim 1 is directed toward a method, which falls within one of the four statutory
categories.
Next, under a Step 2A Prong 1 analysis, the claim recite the following limitations which are interpreted to be, under the broadest reasonable interpretation, abstract ideas:
inserting input test data into the input data stream during operation of the autonomous vehicle (mental process)
The input handler and the checker… are certified to satisfy a risk classification scheme at a predetermined level (mental process)
comparing an output data stream generated by the neural network with a predetermined output corresponding to the input data stream (mental process)
verifying in real-time that the hardware accelerated neural network satisfies the risk classification scheme based on a potential hazard and a corresponding severity, exposure, or controllability of the operating scenario using the input handler and the checker wherein a fault is issued in response to a mismatch between the output data stream and the predetermined output (mental process)
Thus, we have to examine the claim under Step 2A Prong 2, which considers the additional elements with the claim. The claim’s additional elements are:
Generating an input data stream by an input handler
The input data stream comprises sensor data associated with an autonomous vehicle in an operating scenario
The input test data is generated by an input handler responsive to a request for test from a checker that obtains output
A hardware accelerated neural network
The input handler and the checker execute on a first hardware processor
The first hardware processor offloads neural network tasks to hardware that accelerates the neural network
and transmitting the output data stream with the test data removed for use by at least one subsystem of the autonomous vehicle responsive to verifying that the hardware accelerated neural network satisfies the risk classification scheme.
The “input data stream comprises sensor data associated with an autonomous vehicle in an operating scenario” is a limitation that merely indicate the field of use or technological environment and “generally links” sensor data to an autonomous vehicle. (See MPEP2106.05(h)) The “generating an input data stream by an input handler”, “input test data is generated by an input handler responsive to a request for test from a checker that obtains output”, “a hardware accelerated neural network”, “the input handler and the checker execute on a first hardware processor”, and “the first hardware processor offloads neural network tasks to hardware that accelerates the neural network” are limitations that are considered to be mere instructions to apply a judicial exception, as it instructs upon how to use the input handler and checker, as well as the neural network and hardware processor. (See MPEP 2106.05(f)) The limitation, “transmitting the output data stream with the test data removed for use by at least one subsystem of the autonomous vehicle responsive to verifying that the hardware accelerated neural network satisfies the risk classification scheme” is considered to be insignificant extra-solution activity. (See MPEP 2106.05(g)) Therefore, the claim is directed towards an abstract idea.
Under a Step 2B analysis, the claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception as explained above in Step 2A prong 2. Additionally, the limitation, “transmitting the output data stream with the test data removed for use by at least one subsystem of the autonomous vehicle responsive to verifying that the hardware accelerated neural network satisfies the risk classification scheme”, is considered to be well-understood, routine, and conventional, as it is considered receiving or transmitting data over a network. (See MPEP2106.05(d)(ii)) Therefore, the claim is not eligible.
Regarding claim 8, under Step 1 of the Subject Matter Eligibility Test of Products and Processes, claim 8 is directed toward a manufacture, which falls within one of the four statutory categories.
Next, under a Step 2A Prong 1 analysis, the claim recite the following limitations which are interpreted to be, under the broadest reasonable interpretation, abstract ideas:
inserting input test data into the input data stream during operation of the autonomous vehicle (mental process)
The input handler and the checker… are certified to satisfy a risk classification scheme at a predetermined level (mental process)
comparing an output data stream generated by the neural network with a predetermined output corresponding to the input data stream (mental process)
verifying in real-time that the hardware accelerated neural network satisfies the risk classification scheme based on a potential hazard and a corresponding severity, exposure, or controllability of the operating scenario using the input handler and the checker wherein a fault is issued in response to a mismatch between the output data stream and the predetermined output (mental process)
Thus, we have to examine the claim under Step 2A Prong 2, which considers the additional elements with the claim. The claim’s additional elements are:
one program for execution
at least one processor of a first device
generating an input data stream by an input handler
The input data stream comprises sensor data associated with an autonomous vehicle in an operating scenario
The input test data is generated by an input handler responsive to a request for test from a checker that obtains output
A hardware accelerated neural network
The input handler and the checker execute on a first hardware processor
The first hardware processor offloads neural network tasks to hardware that accelerates the neural network
and transmitting the output data stream with the test data removed for use by at least one subsystem of the autonomous vehicle responsive to verifying that the hardware accelerated neural network satisfies the risk classification scheme.
The “input data stream comprises sensor data associated with an autonomous vehicle in an operating scenario” is a limitation that merely indicate the field of use or technological environment and “generally links” sensor data to an autonomous vehicle. (See MPEP2106.05(h)) The “one program for execution”, “one processor of a first device”, “generating an input data stream by an input handler”, “input test data is generated by an input handler responsive to a request for test from a checker that obtains output”, “a hardware accelerated neural network”, “the input handler and the checker execute on a first hardware processor”, and “the first hardware processor offloads neural network tasks to hardware that accelerates the neural network” are limitations that are considered to be mere instructions to apply a judicial exception, as it instructs upon how to use the program, processor, input handler and checker, as well as the neural network and first hardware processor. (See MPEP 2106.05(f)) The limitation, “transmitting the output data stream with the test data removed for use by at least one subsystem of the autonomous vehicle responsive to verifying that the hardware accelerated neural network satisfies the risk classification scheme” is considered to be insignificant extra-solution activity. (See MPEP 2106.05(g)) Therefore, the claim is directed towards an abstract idea.
Under a Step 2B analysis, the claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception as explained above in Step 2A prong 2. Additionally, the limitation, “transmitting the output data stream with the test data removed for use by at least one subsystem of the autonomous vehicle responsive to verifying that the hardware accelerated neural network satisfies the risk classification scheme”, is considered to be well-understood, routine, and conventional, as it is considered receiving or transmitting data over a network. (See MPEP2106.05(d)(ii)) Therefore, the claim is not eligible.
Regarding claim 15, under Step 1 of the Subject Matter Eligibility Test of Products and Processes, claim 15 is directed toward a machine, which falls within one of the four statutory
categories.
Next, under a Step 2A Prong 1 analysis, the claim recite the following limitations which are interpreted to be, under the broadest reasonable interpretation, abstract ideas:
inserting input test data into the input data stream during operation of the autonomous vehicle (mental process)
The input handler and the checker… are certified to satisfy a risk classification scheme at a predetermined level (mental process)
comparing an output data stream generated by the neural network with a predetermined output corresponding to the input data stream (mental process)
verifying in real-time that the hardware accelerated neural network satisfies the risk classification scheme based on a potential hazard and a corresponding severity, exposure, or controllability of the operating scenario using the input handler and the checker wherein a fault is issued in response to a mismatch between the output data stream and the predetermined output (mental process)
Thus, we have to examine the claim under Step 2A Prong 2, which considers the additional elements with the claim. The claim’s additional elements are:
At least one computer-readable medium
At least one processor communicatively coupled the at least one computer-readable medium
generating an input data stream by an input handler
The input data stream comprises sensor data associated with an autonomous vehicle in an operating scenario
The input test data is generated by an input handler responsive to a request for test from a checker that obtains output
A hardware accelerated neural network
The input handler and the checker execute on a first hardware processor
The first hardware processor offloads neural network tasks to hardware that accelerates the neural network
and transmitting the output data stream with the test data removed for use by at least one subsystem of the autonomous vehicle responsive to verifying that the hardware accelerated neural network satisfies the risk classification scheme.
The “input data stream comprises sensor data associated with an autonomous vehicle in an operating scenario” is a limitation that merely indicate the field of use or technological environment and “generally links” sensor data to an autonomous vehicle. (See MPEP2106.05(h)) The “at least one computer-readable medium”, “at least one processor communicatively coupled the at least one computer-readable medium”, “generating an input data stream by an input handler”, “input test data is generated by an input handler responsive to a request for test from a checker that obtains output”, “a hardware accelerated neural network”, “the input handler and the checker execute on a first hardware processor”, and “the first hardware processor offloads neural network tasks to hardware that accelerates the neural network” are limitations that are considered to be mere instructions to apply a judicial exception, as it instructs upon how to use the medium, processor, input handler and checker, as well as the neural network and hardware processor. (See MPEP 2106.05(f)) The limitation, “transmitting the output data stream with the test data removed for use by at least one subsystem of the autonomous vehicle responsive to verifying that the hardware accelerated neural network satisfies the risk classification scheme” is considered to be insignificant extra-solution activity. (See MPEP 2106.05(g)) Therefore, the claim is directed towards an abstract idea.
Under a Step 2B analysis, the claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception as explained above in Step 2A prong 2. Additionally, the limitation, “transmitting the output data stream with the test data removed for use by at least one subsystem of the autonomous vehicle responsive to verifying that the hardware accelerated neural network satisfies the risk classification scheme”, is considered to be well-understood, routine, and conventional, as it is considered receiving or transmitting data over a network. (See MPEP2106.05(d)(ii)) Therefore, the claim is not eligible.
Regarding claims 2, 9, 16, the claims recite “input test data inserted into the input data stream during operation of the autonomous vehicle is static test data generated prior to operation of the autonomous vehicle.” The limitation, as drafted, merely recites the particular technological environment in which the abstract idea takes place, and “generally links” the data being used in the abstract idea to a particular kind of data (static test data). (See MPEP 2106.05(h)) Therefore, the claims are not eligible under U.S.C. 101 for similar reasons as claims 1, 8, and 15.
Regarding claims 3, 10, and 17, the claims recite “input test data inserted into the input data stream during operation of the autonomous vehicle is dynamic test data generated during operation of the autonomous vehicle.” The limitation, as drafted, merely recites the particular technological environment in which the abstract idea takes place, and “generally links” the data being used in the abstract idea to a particular kind of data (dynamic test data). (See MPEP 2106.05(h)) Therefore, the claims are not eligible under U.S.C. 101 for similar reasons as claims 1, 8, and 15.
Regarding claims 4, 11, and 18, the claims recite “inserting input test data into the input data stream during operation of the autonomous vehicle comprises synchronous coordination between the input handler and the checker.” The limitation, as drafted, merely recites the particular technological environment in which the abstract idea takes place, and “generally links” the insertion of input test data used in the abstract idea to data that coordinates with other data synchronously. (See MPEP 2106.05(h)) Therefore, the claims are not eligible under U.S.C. 101 for similar reasons as claims 1, 8, and 15.
Regarding claims 5, 12, 19, the claims recite “inserting input test data into the input data stream during operation of the autonomous vehicle comprises explicit tagging between the input handler and the checker.” The limitation, as drafted, merely recites the particular technological environment in which the abstract idea takes place , and “generally links” the insertion of input test data used in the abstract idea to data that is explicitly tagged. (See MPEP 2106.05(h)) Therefore, the claims are not eligible under U.S.C. 101 for similar reasons as claims 1, 8, and 15.
Regarding claims 6, 13, and 20, the claims recite “input test data is inserted into the input data stream during operation of the autonomous vehicle at predetermined time intervals.” The limitation, as drafted, is interpreted to be mere instructions to apply a judicial exception, as it instructs on when to insert the test data into the input data stream. (See MPEP 2106.05(f)) Therefore, the claims are not eligible under U.S.C. 101 for similar reasons as claims 1, 8, and 15.
Regarding claims 7, 14, and 21, the claims recite “the risk classification scheme is an Automotive Safety Integrity Level (ASIL) defined by International Organization for Standardization (ISO) 26262, and the neural network accelerated by a graphics processing unit is certified at an Automotive Safety Integrity Level.” The limitation, as drafted, merely recites the particular technological environment in which the abstract idea takes place, and “generally links” the neural network accelerated by a graphics processing unit to one certified at an Automotive Safety Integrity Level, and the risk classification scheme being a ASIL defined by the ISO. (See MPEP 2106.05(h)) Therefore, the claims are not eligible under U.S.C. 101 for similar reasons as claims 1, 8, and 15.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claim(s) 1-4, 7-11, 14-18, and 21is/are rejected under 35 U.S.C. 103 as being unpatentable over are rejected under 35 U.S.C. 103 as being unpatentable over Katz et al. (Herein referred to as Katz) (U.S. Patent No. US 11221929 B1) in view of Ghada Bahig and Amr El-Kadi (Herein referred to as Bahig) (Formal Verification of Automotive Design in Compliance With ISO 26262 Design Verification Guidelines) and in further view of Hong et al. (Herein referred to as Hong) (U.S. Patent Application No. US 20210042643 A1)
Regarding claim 1, Katz teaches a method comprising: generating an input data stream by an input handler, wherein the input data stream comprises sensor data associated with an autonomous vehicle in an operating scenario ("In operation, sensor data 968 from one or more sensors are input to the main application processor which functions to generate processed sensor data 972 which is output to the NN processor 966… sensor data 986 from one or more sensors are input to the NN processor 982 which functions to generate processed sensor data 989 and processed insights 988 which is output to the ECU or infotainment system of the vehicle.", column 46, lines 59-62; column 47, lines 26-30) inserting input test data into the input data stream during operation of the autonomous vehicle in the operating scenario, and a hardware accelerated neural network; (“In an alternative embodiment, the RBM co-processor is optionally coupled to the NN device 60 via a suitable interface, e.g., GPUs… In this intralayer safety mechanism, failures in the circuitry along the tensor data flow path (shown in solid lines) within the layer are detected. This is achieved by injecting known tensor test data into the tensor data flow path, calculating an output and comparing that output with a preconfigured expected output.”, column 16, lines 13-16; column 64, lines 62-67; FIG. 70) the first hardware processor offloads neural network tasks to hardware that accelerate the neural network (“The vehicle 940 comprises a plurality of sensors and processors including forward looking camera 942, forward looking radar 944, dashboard camera 949, dashboard lidar 947, mirror camera 943, advanced driver assistant system (ADAS) electronic control unit (ECU) 946, side camera and/or radar 948, display controller 950, vision computer 954, vehicle control computer 952, and drive by wire controller 956.… In this embodiment, the NN processor serves as a dedicated NN accelerator.”, column 46, lines 33-40; column 47, lines 1-2) (One of the plurality of vehicle processors correspond to a first hardware processor that then allocates neural network tasks to neural network processors instead of tasks related to other vehicle functions.) comparing an output data stream generated by the neural network with a predetermined output corresponding to the input data stream; ("...output tensor data is input to the APU which activates the tensor data and utilizes one or more other APU computational resources to generate activation output. This output is compared to expected test output that is calculated and configured in the cluster a priori.", column 65, lines 37-41) and verifying, in real-time that the hardware accelerated neural network satisfies the risk classification scheme based on a potential hazard and severity of the operating scenario (“The major hazards in this configuration, however, are (1) availability failure where the NN processor is unable to perform its designated functionality owing to some system level issue (e.g., cannot power up), meaning that the insights are not received back at the perception subsystem; and (2) false indications where the NN processor is providing wrong output for a given input as either false positives or false negatives… The implications, however, are limited at the platform level since the impact is limited to the subsystem that suffers the impact. In this context, the NN processor acts as a standalone system and thus determines the ASIL level for this subsystem rendering it ASIL-B.”, column 47, lines 10-17; column 47, lines 35-39) (ASIL corresponds to a risk classification scheme.) wherein a fault is issued in response to a mismatch between the output data stream and the predetermined output. ("Several built in self-test (BIST) techniques are employed during runtime operation of the NN processor including powerup BIST, transition BIST, periodic/background BIST, and online/runtime fault detection…This output is compared to expected test output that is calculated and configured in the cluster a priori. In an alternative embodiment, rather than check each and every sample in the output tensor, an ongoing CRC checksum is calculated over the entire test output tensor. Once complete, the CRC checksum is compared to a preconfigured checksum calculated a priori by numeric emulation of the NN processor. Matching CRC checksums are verified and an error flag is raised if a mismatch is found." column 49, lines 33-36; column 65, lines 39-48) (Runtime corresponds to real-time)
However, Katz does not explicitly teach the input test data is generated by an input handler responsive to a request for test from a checker that obtains output of a hardware accelerated neural network, wherein the input handler and the checker execute on a first hardware processor and are certified to satisfy a risk classification scheme at a predetermined level, nor that the verifying, in real-time that the hardware accelerated neural network satisfies the risk classification scheme based on a potential hazard and a corresponding severity, exposure, or controllability of the operating scenario is done specifically by using the input handler and the checker, nor transmitting the output data stream with the test data removed for use by at least one subsystem of the autonomous vehicle responsive to verifying that the hardware accelerated neural network satisfies the risk classification scheme.
Bahig teaches input test data is generated by an input handler responsive to a request for test from a checker that obtains output of a hardware accelerated neural network (“The framework input is a UML state-based system… This input is fed into a model compiler which parses the UML model presented in XML format and constructs object instances of all elements in the UML design. The objects are traversed and mapped into a SAL model and theorems based on transformation rules [9]. SAL objects are also stored and linked to their UML counterparts. Once the SAL model is generated, SAL model checkers get launched to detect any violation against the generated theorems. Any generated counterexample is analyzed by the designer and fixed in UML model domain.”, pg. 4 and 5) (The UML state-based system corresponds to the input handler. The UML model is mapped into a SAL, with the SAL model checkers corresponding to a checker obtaining a SAL model as output.) wherein the input handler and the checker execute on a first hardware processor and are certified to satisfy a risk classification scheme at a predetermined level (“Formal verification of semi-formal model to address: a) Compliance to ISO 26262 test case derivation and software unit design and implementation verification guidelines …Our UML model extensions – satisfiability conditions aim to address the techniques in Figure 2… we initially plan to qualify the model compiler in accordance with ISO 26262 tool qualification guidelines to ensure the safety of the tool and the generated SAL intermediate language.”, pg. 4, right column, under “PROPOSED FRAMEWORK ELEMENTS”; pg. 5, left column, bottom paragraph; pg. 12, left column above “REFERENCES”; See also Figure 2) (ASIL corresponds to the risk classification scheme with the predetermined level corresponding to one of the ASIL levels. (B, C, or D) In combination with the NN processor of Katz, the limitation is taught.) and verifying, in real-time that the hardware accelerated neural network satisfies the risk classification scheme based on a potential hazard and a corresponding severity, exposure, or controllability of the operating scenario using the input handler and the checker, (“Failure of safety critical software could cause hazardous consequences on human life… The first challenge faced was the ability to apply required verification methods… Control flow/data flow analysis were done by manually analyzing the control statements inside source code and creating control flow/data flow graphs for control statements and variables. This was feasible as the modules that were required to be ASIL B compliant were small which rendered this manual effort feasible.”, pg.1, left column, bottom paragraph; pg. 4, left column, second paragraph) (Bahig’s verification method (utilizing the handler and checker) could be easily configured to be performed on Katz neural network, and satisfy a risk classification scheme, (like ASIL) based on the potential hazards outlined in columns 47 and 48 of Katz, and the control flow/data flow analysis of Bahig, which corresponds to a controllability of the operating scenario.)
Therefore, it would have been considered obvious to someone of ordinary skill, prior to the filing date of the current application, to combine the data, the neural network, hardware to operate the neural network, and verification of Katz, with the handler, checker, and specific verification including risk classification of Bahig. One of ordinary skill in art would have been motivated to combined the two teachings, prior to the filing date of the current application, as Bahig’s method presents a framework that allows software designers to verify software in UML that complies with ISO 26262, as disclosed by Bahig. (“present a framework that allows software designers to formally verify a specified software in a semi-formal notation (UML). This complies with ISO 26262 design verification guidelines for ASILs C and D that highly recommend semi-formal verification of the design for ASILs C and D.”, pg. 2, left column, fourth paragraph)
However, the combination does not teach transmitting the output data stream with the test data removed for use by at least one subsystem of the autonomous vehicle responsive to verifying that the hardware accelerated neural network satisfies the risk classification scheme.
Hong teaches transmitting the output data stream with the test data removed for use by at least one subsystem of the autonomous vehicle responsive to verifying that the hardware accelerated neural network satisfies the risk classification scheme. (“In some embodiments, the test data samples 916 to can comprises a subset of the outlier data samples randomly selected and removed from the curated dataset 906 (e.g., wherein the removed test data samples were not used for model updating at 908).” Paragraph 136) (The limitation “for use by at least one subsystem of the autonomous vehicle responsive to verifying that the hardware accelerated neural network satisfies the risk classification scheme”, is given negligible patentable weight, as it is interpreted as “intended use”. Hong teaches test data removed from a set of data, which corresponds to a data stream, which can be easily configured to remove all test data, and work with the output data stream of Katz.)
Therefore, it would have been considered obvious to someone of ordinary skill, prior to the filing date of the current application, to combine the fault detection, verification, and neural network processor of Katz, as modified by Bahig, with the removal of test data as described in Hong. One of ordinary skill in art would have been motivated to combined the teachings, prior to the filing date of the current application, as this allows for the extraction of specific data from learning data in order for a machine learning model to reconstitute learning data to learn more effectively, as disclosed in Hong. (“the learning data reconstitution module 112 may be a module for reconstituting learning data in a case where performance for an artificial intelligence model became lower than or equal to a threshold… The learning data reconstitution module 112 may reconstitute learning data by using a prestored rule. Here, the rule means a rule for an initial artificial intelligence model or a compressed artificial intelligence model to extract specific data from learning data that it learned.”, Paragraph 97 and 98)
Regarding claim 8, Katz teaches A non-transitory computer-readable storage medium comprising at least one program for execution by at least one processor of a first device, ("Any combination of one or more computer usable or computer readable medium(s) may be utilized. The computer-usable or computer-readable medium may be, for example but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, device, or propagation medium.", column 10, lines 21-26) the at least one program including instructions which, when executed by the at least one processor, carry out a method comprising: generating an input data stream by an input handler, wherein the input data stream comprises sensor data associated with an autonomous vehicle in an operating scenario; ("In operation, sensor data 968 from one or more sensors are input to the main application processor which functions to generate processed sensor data 972 which is output to the NN processor 966… sensor data 986 from one or more sensors are input to the NN processor 982 which functions to generate processed sensor data 989 and processed insights 988 which is output to the ECU or infotainment system of the vehicle.", column 46, lines 59-62; column 47, lines 26-30) inserting input test data into the input data stream during operation of the autonomous vehicle in the operating scenario, and a hardware accelerated neural network; (“In an alternative embodiment, the RBM co-processor is optionally coupled to the NN device 60 via a suitable interface, e.g., GPUs… In this intralayer safety mechanism, failures in the circuitry along the tensor data flow path (shown in solid lines) within the layer are detected. This is achieved by injecting known tensor test data into the tensor data flow path, calculating an output and comparing that output with a preconfigured expected output.”, column 16, lines 13-16; column 64, lines 62-67; FIG. 70) the first hardware processor offloads neural network tasks to hardware that accelerate the neural network (“The vehicle 940 comprises a plurality of sensors and processors including forward looking camera 942, forward looking radar 944, dashboard camera 949, dashboard lidar 947, mirror camera 943, advanced driver assistant system (ADAS) electronic control unit (ECU) 946, side camera and/or radar 948, display controller 950, vision computer 954, vehicle control computer 952, and drive by wire controller 956.… In this embodiment, the NN processor serves as a dedicated NN accelerator.”, column 46, lines 33-40; column 47, lines 1-2) (One of the plurality of vehicle processors correspond to a first hardware processor that then allocates neural network tasks to neural network processors instead of tasks related to other vehicle functions.) comparing an output data stream generated by the neural network with a predetermined output corresponding to the input data stream; ("...output tensor data is input to the APU which activates the tensor data and utilizes one or more other APU computational resources to generate activation output. This output is compared to expected test output that is calculated and configured in the cluster a priori.", column 65, lines 37-41) and verifying, in real-time that the hardware accelerated neural network satisfies the risk classification scheme based on a potential hazard and severity of the operating scenario (“The major hazards in this configuration, however, are (1) availability failure where the NN processor is unable to perform its designated functionality owing to some system level issue (e.g., cannot power up), meaning that the insights are not received back at the perception subsystem; and (2) false indications where the NN processor is providing wrong output for a given input as either false positives or false negatives… The implications, however, are limited at the platform level since the impact is limited to the subsystem that suffers the impact. In this context, the NN processor acts as a standalone system and thus determines the ASIL level for this subsystem rendering it ASIL-B.”, column 47, lines 10-17; column 47, lines 35-39) (ASIL corresponds to a risk classification scheme.) wherein a fault is issued in response to a mismatch between the output data stream and the predetermined output. ("Several built in self-test (BIST) techniques are employed during runtime operation of the NN processor including powerup BIST, transition BIST, periodic/background BIST, and online/runtime fault detection…This output is compared to expected test output that is calculated and configured in the cluster a priori. In an alternative embodiment, rather than check each and every sample in the output tensor, an ongoing CRC checksum is calculated over the entire test output tensor. Once complete, the CRC checksum is compared to a preconfigured checksum calculated a priori by numeric emulation of the NN processor. Matching CRC checksums are verified and an error flag is raised if a mismatch is found." column 49, lines 33-36; column 65, lines 39-48) (Runtime corresponds to real-time)
However, Katz does not explicitly teach the input test data is generated by an input handler responsive to a request for test from a checker that obtains output of a hardware accelerated neural network, wherein the input handler and the checker execute on a first hardware processor and are certified to satisfy a risk classification scheme at a predetermined level, nor that the verifying, in real-time that the hardware accelerated neural network satisfies the risk classification scheme based on a potential hazard and a corresponding severity, exposure, or controllability of the operating scenario is done specifically by using the input handler and the checker, nor transmitting the output data stream with the test data removed for use by at least one subsystem of the autonomous vehicle responsive to verifying that the hardware accelerated neural network satisfies the risk classification scheme.
Bahig teaches input test data is generated by an input handler responsive to a request for test from a checker that obtains output of a hardware accelerated neural network (“The framework input is a UML state-based system… This input is fed into a model compiler which parses the UML model presented in XML format and constructs object instances of all elements in the UML design. The objects are traversed and mapped into a SAL model and theorems based on transformation rules [9]. SAL objects are also stored and linked to their UML counterparts. Once the SAL model is generated, SAL model checkers get launched to detect any violation against the generated theorems. Any generated counterexample is analyzed by the designer and fixed in UML model domain.”, pg. 4 and 5) (The UML state-based system corresponds to the input handler. The UML model is mapped into a SAL, with the SAL model checkers corresponding to a checker obtaining a SAL model as output.) wherein the input handler and the checker execute on a first hardware processor and are certified to satisfy a risk classification scheme at a predetermined level (“Formal verification of semi-formal model to address: a) Compliance to ISO 26262 test case derivation and software unit design and implementation verification guidelines …Our UML model extensions – satisfiability conditions aim to address the techniques in Figure 2… we initially plan to qualify the model compiler in accordance with ISO 26262 tool qualification guidelines to ensure the safety of the tool and the generated SAL intermediate language.”, pg. 4, right column, under “PROPOSED FRAMEWORK ELEMENTS”; pg. 5, left column, bottom paragraph; pg. 12, left column above “REFERENCES”; See also Figure 2) (ASIL corresponds to the risk classification scheme with the predetermined level corresponding to one of the ASIL levels. (B, C, or D) In combination with the NN processor of Katz, the limitation is taught.) and verifying, in real-time that the hardware accelerated neural network satisfies the risk classification scheme based on a potential hazard and a corresponding severity, exposure, or controllability of the operating scenario using the input handler and the checker, (“Failure of safety critical software could cause hazardous consequences on human life… The first challenge faced was the ability to apply required verification methods… Control flow/data flow analysis were done by manually analyzing the control statements inside source code and creating control flow/data flow graphs for control statements and variables. This was feasible as the modules that were required to be ASIL B compliant were small which rendered this manual effort feasible.”, pg.1, left column, bottom paragraph; pg. 4, left column, second paragraph) (Bahig’s verification method (utilizing the handler and checker) could be easily configured to be performed on Katz neural network, and satisfy a risk classification scheme, (like ASIL) based on the potential hazards outlined in columns 47 and 48 of Katz, and the control flow/data flow analysis of Bahig, which corresponds to a controllability of the operating scenario.)
Therefore, it would have been considered obvious to someone of ordinary skill, prior to the filing date of the current application, to combine the data, the neural network, hardware to operate the neural network, and verification of Katz, with the handler, checker, and specific verification including risk classification of Bahig. One of ordinary skill in art would have been motivated to combined the two teachings, prior to the filing date of the current application, as Bahig’s method presents a framework that allows software designers to verify software in UML that complies with ISO 26262, as disclosed by Bahig. (“present a framework that allows software designers to formally verify a specified software in a semi-formal notation (UML). This complies with ISO 26262 design verification guidelines for ASILs C and D that highly recommend semi-formal verification of the design for ASILs C and D.”, pg. 2, left column, fourth paragraph)
However, the combination does not teach transmitting the output data stream with the test data removed for use by at least one subsystem of the autonomous vehicle responsive to verifying that the hardware accelerated neural network satisfies the risk classification scheme.
Hong teaches transmitting the output data stream with the test data removed for use by at least one subsystem of the autonomous vehicle responsive to verifying that the hardware accelerated neural network satisfies the risk classification scheme. (“In some embodiments, the test data samples 916 to can comprises a subset of the outlier data samples randomly selected and removed from the curated dataset 906 (e.g., wherein the removed test data samples were not used for model updating at 908).” Paragraph 136) (The limitation “for use by at least one subsystem of the autonomous vehicle responsive to verifying that the hardware accelerated neural network satisfies the risk classification scheme”, is given negligible patentable weight, as it is interpreted as “intended use”. Hong teaches test data removed from a set of data, which corresponds to a data stream, which can be easily configured to remove all test data, and work with the output data stream of Katz.)
Therefore, it would have been considered obvious to someone of ordinary skill, prior to the filing date of the current application, to combine the fault detection, verification, and neural network processor of Katz, as modified by Bahig, with the removal of test data as described in Hong. One of ordinary skill in art would have been motivated to combined the teachings, prior to the filing date of the current application, as this allows for the extraction of specific data from learning data in order for a machine learning model to reconstitute learning data to learn more effectively, as disclosed in Hong. (“the learning data reconstitution module 112 may be a module for reconstituting learning data in a case where performance for an artificial intelligence model became lower than or equal to a threshold… The learning data reconstitution module 112 may reconstitute learning data by using a prestored rule. Here, the rule means a rule for an initial artificial intelligence model or a compressed artificial intelligence model to extract specific data from learning data that it learned.”, Paragraph 97 and 98)
Regarding claim 15, Katz teaches a vehicle (“A diagram illustrating an example vehicle with sensors and related multiple neural network processors is shown in FIG. 40. The vehicle 940 comprises a plurality of sensors and processors including forward looking camera 942, forward looking radar 944, dashboard camera 949, dashboard lidar 947, mirror camera 943, advanced driver assistant system (ADAS) electronic control unit (ECU) 946, side camera and/or radar 948, display controller 950, vision computer 954, vehicle control computer 952, and drive by wire controller 956", column 46, lines 33-40) comprising: at least one computer-readable medium storing computer-executable instructions; (“These computer program instructions may also be stored in a computer-readable medium...”, column 11, lines 15-16) at least one processor communicatively coupled the at least one computer-readable medium and configured to execute the computer executable instructions, the execution carrying out operations, ("In operation, sensor data 968 from one or more sensors are input to the main application processor which functions to generate processed sensor data 972 which is output to the NN processor 966.", column 46, lines 59-62) including generating an input data stream by an input handler, wherein the input data stream comprises sensor data associated with an autonomous vehicle in an operating scenario; ("In operation, sensor data 968 from one or more sensors are input to the main application processor which functions to generate processed sensor data 972 which is output to the NN processor 966… sensor data 986 from one or more sensors are input to the NN processor 982 which functions to generate processed sensor data 989 and processed insights 988 which is output to the ECU or infotainment system of the vehicle.", column 46, lines 59-62; column 47, lines 26-30) inserting input test data into the input data stream during operation of the autonomous vehicle in the operating scenario, and a hardware accelerated neural network; (“In an alternative embodiment, the RBM co-processor is optionally coupled to the NN device 60 via a suitable interface, e.g., GPUs… In this intralayer safety mechanism, failures in the circuitry along the tensor data flow path (shown in solid lines) within the layer are detected. This is achieved by injecting known tensor test data into the tensor data flow path, calculating an output and comparing that output with a preconfigured expected output.”, column 16, lines 13-16; column 64, lines 62-67; FIG. 70) the first hardware processor offloads neural network tasks to hardware that accelerate the neural network (“The vehicle 940 comprises a plurality of sensors and processors including forward looking camera 942, forward looking radar 944, dashboard camera 949, dashboard lidar 947, mirror camera 943, advanced driver assistant system (ADAS) electronic control unit (ECU) 946, side camera and/or radar 948, display controller 950, vision computer 954, vehicle control computer 952, and drive by wire controller 956.… In this embodiment, the NN processor serves as a dedicated NN accelerator.”, column 46, lines 33-40; column 47, lines 1-2) (One of the plurality of vehicle processors correspond to a first hardware processor that then allocates neural network tasks to neural network processors instead of tasks related to other vehicle functions.) comparing an output data stream generated by the neural network with a predetermined output corresponding to the input data stream; ("...output tensor data is input to the APU which activates the tensor data and utilizes one or more other APU computational resources to generate activation output. This output is compared to expected test output that is calculated and configured in the cluster a priori.", column 65, lines 37-41) and verifying, in real-time that the hardware accelerated neural network satisfies the risk classification scheme based on a potential hazard and severity of the operating scenario (“The major hazards in this configuration, however, are (1) availability failure where the NN processor is unable to perform its designated functionality owing to some system level issue (e.g., cannot power up), meaning that the insights are not received back at the perception subsystem; and (2) false indications where the NN processor is providing wrong output for a given input as either false positives or false negatives… The implications, however, are limited at the platform level since the impact is limited to the subsystem that suffers the impact. In this context, the NN processor acts as a standalone system and thus determines the ASIL level for this subsystem rendering it ASIL-B.”, column 47, lines 10-17; column 47, lines 35-39) (ASIL corresponds to a risk classification scheme.) wherein a fault is issued in response to a mismatch between the output data stream and the predetermined output. ("Several built in self-test (BIST) techniques are employed during runtime operation of the NN processor including powerup BIST, transition BIST, periodic/background BIST, and online/runtime fault detection…This output is compared to expected test output that is calculated and configured in the cluster a priori. In an alternative embodiment, rather than check each and every sample in the output tensor, an ongoing CRC checksum is calculated over the entire test output tensor. Once complete, the CRC checksum is compared to a preconfigured checksum calculated a priori by numeric emulation of the NN processor. Matching CRC checksums are verified and an error flag is raised if a mismatch is found." column 49, lines 33-36; column 65, lines 39-48) (Runtime corresponds to real-time)
However, Katz does not explicitly teach the input test data is generated by an input handler responsive to a request for test from a checker that obtains output of a hardware accelerated neural network, wherein the input handler and the checker execute on a first hardware processor and are certified to satisfy a risk classification scheme at a predetermined level, nor that the verifying, in real-time that the hardware accelerated neural network satisfies the risk classification scheme based on a potential hazard and a corresponding severity, exposure, or controllability of the operating scenario is done specifically by using the input handler and the checker, nor transmitting the output data stream with the test data removed for use by at least one subsystem of the autonomous vehicle responsive to verifying that the hardware accelerated neural network satisfies the risk classification scheme.
Bahig teaches input test data is generated by an input handler responsive to a request for test from a checker that obtains output of a hardware accelerated neural network (“The framework input is a UML state-based system… This input is fed into a model compiler which parses the UML model presented in XML format and constructs object instances of all elements in the UML design. The objects are traversed and mapped into a SAL model and theorems based on transformation rules [9]. SAL objects are also stored and linked to their UML counterparts. Once the SAL model is generated, SAL model checkers get launched to detect any violation against the generated theorems. Any generated counterexample is analyzed by the designer and fixed in UML model domain.”, pg. 4 and 5) (The UML state-based system corresponds to the input handler. The UML model is mapped into a SAL, with the SAL model checkers corresponding to a checker obtaining a SAL model as output.) wherein the input handler and the checker execute on a first hardware processor and are certified to satisfy a risk classification scheme at a predetermined level (“Formal verification of semi-formal model to address: a) Compliance to ISO 26262 test case derivation and software unit design and implementation verification guidelines …Our UML model extensions – satisfiability conditions aim to address the techniques in Figure 2… we initially plan to qualify the model compiler in accordance with ISO 26262 tool qualification guidelines to ensure the safety of the tool and the generated SAL intermediate language.”, pg. 4, right column, under “PROPOSED FRAMEWORK ELEMENTS”; pg. 5, left column, bottom paragraph; pg. 12, left column above “REFERENCES”; See also Figure 2) (ASIL corresponds to the risk classification scheme with the predetermined level corresponding to one of the ASIL levels. (B, C, or D) In combination with the NN processor of Katz, the limitation is taught.) and verifying, in real-time that the hardware accelerated neural network satisfies the risk classification scheme based on a potential hazard and a corresponding severity, exposure, or controllability of the operating scenario using the input handler and the checker, (“Failure of safety critical software could cause hazardous consequences on human life… The first challenge faced was the ability to apply required verification methods… Control flow/data flow analysis were done by manually analyzing the control statements inside source code and creating control flow/data flow graphs for control statements and variables. This was feasible as the modules that were required to be ASIL B compliant were small which rendered this manual effort feasible.”, pg.1, left column, bottom paragraph; pg. 4, left column, second paragraph) (Bahig’s verification method (utilizing the handler and checker) could be easily configured to be performed on Katz neural network, and satisfy a risk classification scheme, (like ASIL) based on the potential hazards outlined in columns 47 and 48 of Katz, and the control flow/data flow analysis of Bahig, which corresponds to a controllability of the operating scenario.)
Therefore, it would have been considered obvious to someone of ordinary skill, prior to the filing date of the current application, to combine the data, the neural network, hardware to operate the neural network, and verification of Katz, with the handler, checker, and specific verification including risk classification of Bahig. One of ordinary skill in art would have been motivated to combined the two teachings, prior to the filing date of the current application, as Bahig’s method presents a framework that allows software designers to verify software in UML that complies with ISO 26262, as disclosed by Bahig. (“present a framework that allows software designers to formally verify a specified software in a semi-formal notation (UML). This complies with ISO 26262 design verification guidelines for ASILs C and D that highly recommend semi-formal verification of the design for ASILs C and D.”, pg. 2, left column, fourth paragraph)
However, the combination does not teach transmitting the output data stream with the test data removed for use by at least one subsystem of the autonomous vehicle responsive to verifying that the hardware accelerated neural network satisfies the risk classification scheme.
Hong teaches transmitting the output data stream with the test data removed for use by at least one subsystem of the autonomous vehicle responsive to verifying that the hardware accelerated neural network satisfies the risk classification scheme. (“In some embodiments, the test data samples 916 to can comprises a subset of the outlier data samples randomly selected and removed from the curated dataset 906 (e.g., wherein the removed test data samples were not used for model updating at 908).” Paragraph 136) (The limitation “for use by at least one subsystem of the autonomous vehicle responsive to verifying that the hardware accelerated neural network satisfies the risk classification scheme”, is given negligible patentable weight, as it is interpreted as “intended use”. Hong teaches test data removed from a set of data, which corresponds to a data stream, which can be easily configured to remove all test data, and work with the output data stream of Katz.)
Therefore, it would have been considered obvious to someone of ordinary skill, prior to the filing date of the current application, to combine the fault detection, verification, and neural network processor of Katz, as modified by Bahig, with the removal of test data as described in Hong. One of ordinary skill in art would have been motivated to combined the teachings, prior to the filing date of the current application, as this allows for the extraction of specific data from learning data in order for a machine learning model to reconstitute learning data to learn more effectively, as disclosed in Hong. (“the learning data reconstitution module 112 may be a module for reconstituting learning data in a case where performance for an artificial intelligence model became lower than or equal to a threshold… The learning data reconstitution module 112 may reconstitute learning data by using a prestored rule. Here, the rule means a rule for an initial artificial intelligence model or a compressed artificial intelligence model to extract specific data from learning data that it learned.”, Paragraph 97 and 98)
Regarding claims 2, 9, 16, Katz, as modified by Bahig and Hong, teaches the method, the non-transitory computer readable storage medium, and vehicle of claims 1, 8, and 15 respectively wherein input test data inserted into the input data stream during operation of the autonomous vehicle is static test data generated prior to operation of the autonomous vehicle. (The test data may be provided by one of several sources: (1) test data 1484 stored in L3 memory 1482; (2) test data 1486 stored in a register in the cluster or elsewhere; and (3) test data (and optionally weights) generated dynamically on the fly via a test data generator 1488.”, column 65, lines 57-62 (Katz))
Regarding claims 3, 10, and 17, Katz, as modified by Bahig and Hong, teaches the method, the non-transitory computer readable storage medium, and vehicle of claims 1, 8, and 15 respectively wherein input test data inserted into the input data stream during operation of the autonomous vehicle is dynamic test data generated during operation of the autonomous vehicle. ("The test data may be provided by one of several sources: (1) test data 1484 stored in L3 memory 1482; (2) test data 1486 stored in a register in the cluster or elsewhere; and (3) test data (and optionally weights) generated dynamically on the fly via a test data generator 1488.", column 65, lines 57-62 (Katz))
Regarding claims 4, 11, and 18, Katz, as modified by Bahig and Hong, teaches the method, the non-transitory computer readable storage medium, and vehicle of claims 1, 8, and 15 respectively wherein inserting input test data into the input data stream during operation of the autonomous vehicle comprises synchronous coordination between the input handler and checker. ("Trigger signals are used to trigger activity. Triggers can be issued to activate other triggers. They represent an asynchronous mechanism that functions to synchronize activities in the NN processor. For example, a trigger can be issued to halt processing until a buffer is written to, or until a layer completes processing (or otherwise function as an indication that some event has taken place and further processing can commence).", column 34, lines 46-53 (Katz)) (The synchronous coordination is performed by trigger signals, and synchronize activities in the neural network processor. The input handler and checker of Bahig can be easily configured to run inside the NN processor of Katz to teach this limitation.)
Regarding claims 7, 14, and 21, Katz, as modified by Bahig and Hong, teaches the method, the non-transitory computer readable storage medium, and vehicle of claims 1, 8, and 15 respectively wherein the risk classification scheme is an Automotive Safety Integrity Level (ASIL) defined by International Organization for Standardization (ISO) 26262 (“An automotive functional safety standard, ISO 26262 [3], has been published in 2011 whose objectives are: providing an automotive safety lifecycle (management, development, production, operation, service, decommissioning), supports tailoring the necessary activities during these lifecycle phases, and providing an automotive specific risk-based approach for determining risk classes (Automotive Safety Integrity Levels, ASILs… we will present a framework that allows software designers to formally verify a specified software in a semi-formal notation (UML). This complies with ISO 26262 design verification guidelines for ASILs C and D that highly recommend semi-formal verification of the design for ASILs C and D.”, pgs. 1-2; pg. 2, left column, fourth paragraph) (Bahig)) and the neural network accelerated by the graphics processing unit is certified at an Automotive Safety Integrity Level. ("the NN processor may be part of an automotive safety integrity level (ASIL) D design and will thus have system level redundancy.", column 47, lines 7-9 (Katz))
Claim(s) 5, 12, and 19 are rejected under 35 U.S.C. 103 as being unpatentable over Katz in view of Bahig, in further view of Hong, and in further view of Simoundis. (U.S. Patent Application No. US 20200364953 A1)
Regarding claims 5, 12, and 19, Katz, as modified by Bahig and Hong, teaches the method, non-transitory computer-readable medium, and vehicle of claims 1, 8, and 15 respectively. However, Katz does not teach explicit tagging between the input handler and checker.
Simoundis teaches explicit tagging between the input handler and checker. ("Data processing may include, for example, data normalization, labeling data with metadata, tagging, data alignment, data segmentation, and various others.", Paragraph 143) (Processed data can be easily configured to be handled between the input handler and checker, with evidence supporting as much in Bahig. “The requirements are mapped into UML packages, components, classes (attributes and operations), and state machines. All defined data types, attributes, functions are defined in the UML model. Once UML model is complete, the model currently captures architectural design of the specification. OAL (Object Action Language) is now embedded in states, transitions, operations (Instance or class based), ports, mathematically derived attributes, and functions to capture the specification behavior.”, pgs. 4-5)
Therefore, it would have been considered obvious to someone of ordinary skill, prior to the filing date of the current application, to combine the method/non-transitory computer-readable medium/vehicle of Katz, modify it with Bahig and Hong, and add the data tagging of Simoundis. One of ordinary skill in the art would have been motivated to do so, as data tagging can be used in multiple ways when it comes to the management of autonomous vehicle data. It can be used in: repository storage, data processing/normalization, annotating multimedia data, and third-party functions that incorporate data tagging, as disclosed by Simoundis. (“the data processing module 411 may support ingesting of sensor data into a local storage repository (e.g., local time-series database), data cleansing, data enrichment (e.g., decorating data with metadata), data alignment, data annotation, data tagging, data aggregation, and various other data processing”, “Data processing may include, for example, data normalization, labeling data with metadata, tagging, data alignment, data segmentation, and various others”, “…the plurality of functions may comprise third-party functions such as ingestion 901, filtering 905, cleaning 907, tagging 909, augmentation 911, annotation 913, anonymization 915, and various others (e.g., simulate)… In a further example, data tagging 909 or annotation 913 may include annotation of multimedia data (e.g., image, Lidar, audio) that happens at every level and creation of metadata.”, Paragraphs 116, 143, and 147)
Claim(s) 6, 13, and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Katz in view of Bahig, in further view of Hong, and in further view of Boliang et al. (Herein referred to as Boliang) (U.S. Patent Application No. US 20210122383 A1)
Regarding claims 6, 13, and 20, Katz, as modified by Bahig and Hong, teaches the method, non-transitory computer-readable medium, and vehicle of claims 1, 8, and 15 respectively. However, Katz does not teach input test data is insert into the input data stream during operation of the autonomous vehicle at predetermined time intervals.
Boliang teaches input test data is insert into the input data stream during operation of the autonomous vehicle at predetermined time intervals. ("it is possible to store only the test data that are relevant within a driving scenario. These may be, in particular, test data within a time interval before or after an accident or after an error in the driving program has occurred.”, Paragraph 20)
Therefore, it would have been considered obvious to someone of ordinary skill, prior to the filing date of the current application, to combine the method/non-transitory computer-readable medium/vehicle of Katz, as modified by Bahig and Hong, with the test data within particular time intervals of Boliang. One would be motivated to combine the teachings, as storing only relevant test data relevant to a driving scenario during a predefined period of time reduces the amount of data needed to be stored, as disclosed in Boliang. (“According to one further specific embodiment of the present invention, at least a portion of the control commands of the driving program for the test vehicle are detected and stored during a predefined period of time. To further reduce stored data, it is possible to store only the test data that are relevant within a driving scenario.”, Paragraph 20)
Response to Arguments
Applicant's arguments filed on March 30th, 2026 have been fully considered but they are not fully persuasive. The applicant’s amendments are enough to overcome the 112 rejection, and so the 112 rejections have been withdrawn. The applicant argues in substance:
Argument 1: The claim as a whole recites improvements in collecting traffic data rather than an improvement of an alleged abstract idea. Example 40 of the 2019 Subject Matter Eligibility Examples applies to this case.
The examiner respectfully disagrees. In Example 40, the claims as a whole recites an improvement in the collection of data, and describes in detail how to data is collected and used. By comparison, the claims in the current application recite not an improvement in data collection, but at most a potential improvement in data processing, with the only step reciting data collection being the “generating” step incorporating sensor data. As such, Example 40’s precedent does not apply to this case. Furthermore, the claim as a whole is not even directed towards an improvement in data processing, but rather interpreted to be an improvement of an abstract idea.
Argument 2: The amended claims recite insertion of test data at runtime which is an improvement over prior systems, As such, the claims recite a practical application of the alleged abstract idea, and are not directed towards an abstract idea.
The examiner respectfully disagrees. As explained above, the insertion of test data during runtime does not recite an improvement over prior systems, but rather is interpreted to be, under the broadest reasonable interpretation, a mental process, which is a grouping of abstract idea. As such, it is not considered as an additional element but as part of the abstract idea, and as such does not recite a practical application of the abstract idea. The 101 rejections are maintained.
Argument 3: Katz, as modified by Bahig, does not properly address the limitation: verifying an integrity of in real-time that the neural network accelerated by the graphics processing unit satisfies the risk classification scheme based on the input handler and the checker, wherein the risk classification scheme that identifies components considered safe to implement in the autonomous vehicle. Specifically, Katz and Bahig do not appear to teach or suggest real-time verification of the integrity of a neural network accelerated by a graphics processing unit.
The examiner respectfully disagrees. As established above in this action and in previous actions, Katz teaches verifying an integrity of in real-time that the neural network accelerated by the graphics processing unit satisfies the risk classification scheme… wherein the risk classification scheme identifies components considered safe to implement in the autonomous vehicle. What Katz does not teach is the explicit use of an input handler and checker wherein the verification step is based on an input handler and checker. Bahig remedies the deficiencies and teaches a method of verification using an input handler and checker, as stated above. Therefore, the combination fully teaches the limitation, and the rejection is proper.
Argument 4: Katz, as modified by Bahig, does not teach a risk classification scheme based on a potential hazard and corresponding severity, exposure or controllability.
The examiner respectfully disagrees. As explained above, Katz a risk classification scheme based on a potential hazard and corresponding severity. However, Katz does not explicitly teach the verification step with the use of an input handler and checker. Bahig teaches a risk classification scheme that utilizes an input handler and checker. Therefore, the combination fully teaches the limitation, and the rejection is proper, and the 103 rejections are maintained
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Tyler E Iles whose telephone number is (571)272-5442. The examiner can normally be reached 9:00am - 5:00pm.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Kakali Chaki can be reached at (571) 272-3719. 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.
/T.E.I./ Patent Examiner, Art Unit 2122
/KAKALI CHAKI/ Supervisory Patent Examiner, Art Unit 2122
Prosecution Insights