Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Continued Examination Under 37 CFR 1.114
A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant’s submission filed on February 2nd, 2026 has been entered.
Amendments
This action is in response to amendments filed January 14th, 2026 as part of an after-final amendment, but not entered at that time, in which Claims 1, 8, 15 were amended. The amendments have now been entered, and Claims 1, 3-8, 10-15, and 17-21 are currently pending.
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.
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, 4-6, 8, 11-13, 15, and 18-20 are rejected under 35 U.S.C. 103 as being unpatentable over Wandzik et al., “Uncertainty Quantification in Deep Residual Networks,” in view of He et al., “Identity Mappings in Deep Residual Networks,” and further in view of Yildirim et al., “Leveraging Uncertainty in Deep Learning for Selective Classification,” and Broggi, US Patent 11,117,570.
Regarding Claim 1, Wandzik teaches a system comprising a computer including a processor and a memory, the memory including instructions such that the processor is programmed (Wandzik, pg. 5, Section 4.3 performs their methods on a computer) to: calculate a [uncertainty] of a plurality of predictions of values obtained by inputting sensor data to each of a plurality of deep neural network and a mean of the values, wherein each prediction is generated by a different one of the deep neural network from the sensor data, wherein the different deep neural networks include skip respective skip connections, wherein the respective skip connections in each of the different neural networks skip one or more different layers than the respective skip connections in others of the different deep neural networks (Wandzik, pg.1, Fig. 1 shows a base residual neural network with skip connections where different layers/residual blocks are present in different instantiations of the network, see “performing several stochastic passes through the network while randomly switching residual blocks” that is, networks with different “active/inactive” blocks are a plurality of deep neural networks; image data is sensor data, and “how much the model is confident in its prediction” is a uncertainty; also see pg. 4, 2nd column, Algorithm 1, “MCSD” which produces the mean “Mean prediction” from T different neural networks obtained by randomly sampling the different layers).
Wandzik uses ResNets with skip connections (Wandzik, pg. 5, 2nd column, 2nd paragraph, “We evaluate all methods using the ResNet-100 architecture”) but is silent as to whether the skip connections are output between batch normalization layers and activation layers to be input by one or more later layers of the deep neural network. However, He teaches that this design of a ResNet is an “original Residual Unit” (He, pg. 2, Fig. 1(a), where the skip connection is output after the ReLU/activation layer of one instance of the illustrated “arrow” and before a weight/BN layer of the next instance of that “arrow” which, in accordance with the scope of the claim as interpreted in light of the specification, [0071], “the skip connections are between the convolutional layers/batch normalization layers 202 and the activation layers 306”). It would have been obvious to one of ordinary skill in the art to use the configuration of a ResNet, as illustrated in He, in the ResNet of Wandzik. The motivation to do so is that it is the “original” ResNet design.
Algorithm 1 of Wandzik computes a measure of confidence/uncertainty via “predictive entropy” (see Wandzik, pg. 4, 1st column, last paragraph) and while Wandzik mentions the variance of the number of predictions (Wandzik, pg. 1, Fig. 1 & pg. 5, 1st column, 1st paragraph), Wandzik does not compute standard deviation as a measure of the confidence/uncertainty. However, Yildirim does teach to compute the standard deviation of a number of stochastically determined neural networks as a measure of confidence/uncertainty (Yildirim, pg. 6, 2nd column, “We run trained and optimized DNN 100 times with its dropout open” that is, 100 stochastically determined different neural networks “We use empirical mean and standard deviation of these 100 softmax outputs as the predictive mean and the model uncertainty”). It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to use the standard deviation instead of entropy in calculation of Wandzik. The reason to do so is that standard deviation is presented (in Wandzik and in Yildirim) as an alternative way to measure uncertainty/confidence.
Yildirim further teaches to compare the standard deviation of the plurality of predictions with a predetermined variation threshold (Yildirim, pg. 3, 2nd column, Fig. 1 the vertical axis & Table 1, “Upper uncertainty boundary for positive/negative decisions”); when the standard deviation is less than the predetermined variation threshold, output [a decision] (Yildirim, pg. 3, 2nd column, Fig. 1, “
A
1
” and “
A
4
” & 1st column, last paragraph, “We characterize five decision areas of classification and rejection and graphically demonstrate these areas in Figure 1.
A
1
defines the decision region for positive classification”); when the standard deviation is greater than the predetermined threshold variation [reject the decision as too uncertain] (Yildirim, pg. 3, 1st column, last paragraph, “
A
2
and
A
5
are rejection regions due to their high model uncertainty”).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to apply Yildirim’s rules for accepting or rejecting a decision of a neural network based on uncertainty thresholds to the residual network (with uncertainty calculations) of Wandzik. The motivation to do so is “to identify optimal classification and rejection regions” (Yildirim, Abstract) that is, to decide when to accept the decision of a classifier.
The Wandzik/He/Yildirim combination teaches general method for thresholding rejection regions of too high uncertainty in a neural network classifier system, and thus does not teach to output at least one measurement of an object when the uncertainty is low nor to disable a vehicle mode for controlling vehicle propulsion, breaking, and/or steering when the uncertainty is high. However, Broggi teaches at least one measurement of an object based on the predictions (Broggi, column 41, lines 30-32, “detect the fire hydrant, the tree, and/or the vehicle with a confidence greater than the threshold”) and when the standard deviation is high to disable a vehicle mode controlling vehicle propulsion, braking, and/or steering by actuating vehicle subsystems (Broggi, column 41, lines 30-41, “if the processors detect … with a confidence level greater than the threshold value, then the processors may … enable the vehicle maneuver …. [when] the confidence level based on the computer vision analysis of the example video frame may be below the threshold value, then the processors may not have sufficient information to enable autonomous control” denotes disabling under low confidence/high standard deviation scenarios). It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to apply the Wandzik/Yildirim combination to the object detection and autonomous vehicle control scenario of Broggi. The motivation to do so is that Broggi is an application of a vision analysis system (like that of Wandzik) that needs a confidence value and threshold (like one that Wandzik/Yildirim provides).
Regarding Claim 4, the Wandzik/He/Yildirim/Broggi combination of Claim 1 teaches the system of Claim 1 (and thus the rejection of Claim 1 is incorporated). The combination has already been shown to teach to receive the sensor data from a vehicle sensor of a vehicle (via incorporating the vision system of the vehicle of Broggi) and provide the sensor data to each deep neural network (Wandzik, pg. 4, Algorithm 1).
Regarding Claim 5, the Wandzik/He/Yildirim/Broggi combination of Claim 1 teaches the system of Claim 1 (and thus the rejection of Claim 1 is incorporated). Wandzik further teaches wherein each deep neural network comprises a convolutional neural network (Wandzik, pg. 5, 2nd column, 2nd paragraph, “We evaluate all methods using the ResNet-110 architecture” with pg. 3, 2nd column, last paragraph, “Deep residual networks (ResNet) … employ skip connections, parallel to convolutional layers”).
Regarding Claim 6, the Wandzik/He/Yildirim/Broggi combination of Claim 5 teaches the system of Claim 5 (and thus the rejection of Claim 1 is incorporated). The combination of using Wandzik’s residual neural network to perform image analysis on Broggi’s vehicle further teaches provide an image captured by an image sensor of a vehicle to each convolutional neural network (Wandzik, pg. 1, Fig. 1, where the images are from Broggi’s vehicle); and calculate the plurality of predictions based on the image (Wandzik, pg. 1, Fig. 1 & pg. 4, Algorithm 1).
Claims 8 and 11-13 recite a system comprising: a server; and a vehicle including a vehicle system, the vehicle system comprising precisely the computers of Claims 1 and 4-6. As Broggi implement their system on a vehicle that may include a server (Broggi, column 52, lines 28-29, “The devices may include … servers” & Fig. 1, Processor includes CNN 150 on the vehicle), Claims 15 and 18-20 are rejected for reasons set forth in the rejections of Claims 1 and 4-6, respectively.
Claims 15 and 18-20 recite precisely the method that the computer of Claims 1 and 4-6, respectively, are configured to perform, and are thus rejected for reasons set forth in the rejections of Claims 1 and 4-6, respectively.
Claims 3, 10, and 17 are rejected under 35 U.S.C. 103 as being unpatentable over Wandzik, in view of He, Yildirim, and Broggi, and further in view of Lin, US PG Pub 2020/0174493.
Regarding Claim 3, the Wandzik/He/Yildirim/Broggi combination of Claim 1 teaches the system of Claim 1 (and thus the rejection of Claim 1 is incorporated). The combination does not teach when the standard deviation is high to transmit the sensor data to a server, but Lin teaches, regarding images taken from vehicle camera, to transmit, to a server, the sensor data when [a trigger circumstance occurs] (Lin, [0010], “vehicles may continuously or periodically transmit data to other vehicles or infrastructure (e.g. the central server), and/or the vehicles may transmit data in response to a triggering event”). Transmitting continuously teaches transmit the sensor data when the standard deviation is greater than the predetermined variation threshold as well as to transmit when the deviation is less that the threshold. Further, loss of an ability to make a prediction because the standard deviation is above the threshold is also a possible triggering event. It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to either transmit data of the combination continuously or to transmit data when the threshold triggers a transmitting event, to a server, as does Lin. The motivation to do so is share data among different entities (Lin, [0060] & Fig. 2).
Claim 10 recites a system comprising: a server; and a vehicle including a vehicle system, the vehicle system comprising precisely the computer of Claim 3. As Broggi teaches that their processors may be a server (Broggi, column 52, lines 28-29, “The devices may include … servers”) and that their processors that process the CNN reside on a vehicle (Broggi, Fig. 1), Claim10 is thus rejected for reasons set forth in the rejection of Claim 3.
Claim 17 recites precisely the method that the computer of Claim 3 is configured to perform, and is thus rejected for reasons set forth in the rejection of Claim 3.
Claims 7, 14, and 21 are rejected under 35 U.S.C. 103 as being unpatentable over Wandzik, in view of He, Yildirim, and Broggi, and further in view of Dahal, “DeepTrailerAssist: Deep Learning based trailer detection, tracking, and articulation estimation on automotive rear-view camera.”
Regarding Claim 7, the Wandzik/He/Yildirim/Broggi combination of Claim 1 teaches the system of Claim 1 (and thus the rejection of Claim 1 is incorporated). The combination does not teach wherein the object comprises at least a portion of a trailer connected to a vehicle and the measurement comprises a trailer angle, but Dahal teaches to determine these things as part of an autonomous control mode of a vehicle (Dahal, pg. 5, Fig. 6). It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to apply the Wandzik/Yildirim/Broggi invention (which ensures that a vehicle has confident enough predictions from a vision system to enable or disable an autonomous action of the vehicle) to the autonomous parking system of Dahal. The motivation to do so is because trailer control is a vision system of a vehicle that can autonomously be controlled and which desires confident knowledge of the environment (Dahal, pg. 6, 1st column, 1st paragraph).
Claim 14 recites a system comprising: a server; and a vehicle including a vehicle system, the vehicle system comprising precisely the computer of Claim 7. As Broggi teaches that their processors may be a server (Broggi, column 52, lines 28-29, “The devices may include … servers”) and that their processors that process the CNN reside on a vehicle (Broggi, Fig. 1), Claim 14 is thus rejected for reasons set forth in the rejection of Claim 7.
Claim 21 recites precisely the method that the computer of Claim 7 is configured to perform, and is thus rejected for reasons set forth in the rejection of Claim 7.
Response to Arguments
Applicant’s arguments filed January 14th, 2026 have been fully considered, but are not fully persuasive.
Applicant’s arguments regarding the 35 U.S.C. 103 rejections of the previous office action are moot, as they do not refer to new reference He being used to teach the amended limitation. Specifically, He teaches a ResNet which outputs a skip connection from a point in a ResNet after an activation (ReLU) layer and before a batch normalization layer, i.e. between them.
Applicant’s arguments regarding the other independent claims and dependent claims rely upon those argued with respect to Claim 1, and are thus unpersuasive.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to BRIAN M SMITH whose telephone number is (469)295-9104. The examiner can normally be reached Monday - Friday, 8:00am - 4pm Pacific.
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.
/BRIAN M SMITH/Primary Examiner, Art Unit 2122
Prosecution Insights