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 .
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and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114.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 05/07/2026 has been entered.
Response to Amendment
The office action is in response to Applicant’s amendment filed 05/07/2026 which has been entered and made of record. Claims 1, 15 and 18 have been amended. No claim has been newly added. Claims 1-20 are pending in the application.
Response to Arguments
Applicant’s arguments, filed 05/07/2026, with respect to the rejection(s) under 35 U.S.C. 103 have been fully considered and are persuasive. Therefore, the rejection has been withdrawn. However, upon further consideration, a new ground(s) of rejection is made in view of Anastassov and Cho as fully explained below.
Applicant argues, Anastassov, Liu and Kaku taken individually or in combination, do not teach the newly amended limitations of independent claims.
Examiner agrees. However, upon further consideration, a new ground(s) of rejection is made in view of Anastassov and Cho as fully explained below.
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, 13-15, 17-18 and 20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Anastassov (US 20220161817 A1), hereinafter as Anastassov, in view of Cho et al. (US 20260109373 A1), hereinafter as Cho.
Regarding claim 1, Anastassov teaches A method (Anastassov paragraph [0007] “a method comprises retrieving probe data collected from one or more sensors of one or more probe devices traveling within a geographic area including at least one geographic partition.”) comprising: generating, for each cell of one or more cells of a grid representing a region of an environment (Anastassov paragraph [0038] “FIG. 2A is a diagram illustrating an example geo-spatial partitioning scheme, according to one embodiment. The diagram 200 shows an area of interest (e.g., a Washington D.C. map 201) divided into partitions 203 for the purposes of distributed/parallelized processing……By way of example, a partition 203m containing a portion of Washington D.C. borderline 201n is divided into grid cells 205”), a cell representation indicating one or more points that correspond with the cell and that represent corresponding sensor detections generated by one or more ego-machines in the environment (Anastassov teaches the probe data as the cell representation, paragraph [0008] “cause the apparatus to retrieve probe data collected from one or more sensors of one or more probe devices traveling within a geographic area including at least one geographic partition.” And paragraph [0036] “the system 100 can prepare probe data into seed points, then continue with the seed points (without the probe data)”); …… and generating, based at least on the lane data, a lane graph that represents the one or more lanes on one or more roads in the environment (Anastassov paragraph [0066] “the seed points 225 created in step 407 (e.g., FIGS. 2B-2C) can be used to create continuous road paths represented by polylines 1-6 in FIG. 2D.”, and paragraph [0082] “In one embodiment, in step 411, the output module 309 can include the at least one continuous road path in an output representing a base map of the geographic area. In other embodiments, the base map further includes connections, intersections, splits/merges, etc. as a graph.”).
Anastassov does not explicitly teach applying the cell representation for at least one of the one or more cells to one or more decoders of one or more neural networks; generating, as output of the one or more decoders, lane data indicating one or more lanes associated with the one or more cells, one or more cross-sections of the one or more lanes associated with the one or more cells, and one or more connections between the one or more cross-sections;
Cho teaches applying the cell representation for at least one of the one or more cells to one or more decoders of one or more neural networks (Cho Figure 2, paragraph [0024] “With the output set of tensors, the computing system on the ego can apply a third set of encoders (e.g., an autoregressive decoder) to generate a set of tokens for the graph. Each token can define various properties of a point forming one or more of the lane segments through the environment.”, paragraph [0074-0075] “The tensor 250 can include the aggregate set of embeddings derived from both the sensor data and the map data …… The autoregressive decoder of the lane encoder 260 can include a sequence of models to process portions of input embeddings from the tensor 250 to generate output embeddings dependent on another output embedding derived from prior portion of input embeddings.”); generating, as output of the one or more decoders, lane data indicating one or more lanes associated with the one or more cells, one or more cross-sections of the one or more lanes associated with the one or more cells, and one or more connections between the one or more cross-sections (Cho teaches output tokens from machine learning model, further teaches tokens with point property and connectivity information. Cho paragraph [0063] “The output can include environment features (e.g., attributes gathered from the sensor data), map features (e.g., attributes in navigation map such as topological features and road layouts), classifications (e.g., a type of topology), and an output token (e.g., a combination of environment features, map features, and classifications) to be included in a graph defining lane segments”, paragraph [0075] “From processing the tensor 250, the lane encoder 260 can produce, output, or otherwise generate a set of lane instances 265 and at least one adjacency matrix 270. The lane instances 265 can define one or more lane segments through which the ego can potentially navigate the environment. The adjacent matrix 270 can define a connectivity or relationship among the lane segments defined in the lane segments 265.” And paragraph [0024] “Each token can define various properties of a point forming one or more of the lane segments through the environment ……. the computing system can classify the point by a type of topology, such as a starting point, a continuation, a fork, or a terminal point”);
Anastassov and Cho are in the same field of endeavor, namely computer graphics, especially in the field of road map generation based on sensor data. Cho teaches using an AI model with decoder to improve accuracy (Cho paragraph [0029] “the data ingested and/or predicted by the AI model(s) 110c with respect to the egos 140 (at inference time) may also be used to improve the AI model(s) 110c. Therefore, the system 100 depicts a continuous loop that can periodically improve the accuracy of the AI model(s) 110c. Moreover, the system 100 depicts a loop in which data received the egos 140 can be used to at training phase in addition to the inference phase.”). Therefore, it would have been obvious for a person of ordinary skill in the art before the effective filing date of the claimed invention to combine the teaching of Cho with the method of Anastassov to achieve accuracy and efficiency.
Regarding claim 13, Anastassov in view of Cho teach The method of claim 1, and further teach wherein each cell representation comprises a set of point representations including a set of attributes values for attributes associated with the one or more points that correspond with the cell (Anastassov paragraph [0047] “the sensor data includes probe data may be reported as probes, which are individual data records collected at a point in time that records telemetry data for that point in time. A probe point can include attributes such as: (1) source ID, (2) longitude, (3) latitude, (4) elevation, (5) heading, (6) speed, (7) time, and (8) access type. A source/probe can be a vehicle, a drone, a user device travelling with the vehicle, etc. Probe data can be used to define probe (e.g., a vehicle) travel paths, count numbers of contributing vehicles, forming “drives” by a location point (together with time information), etc.”).
Regarding claim 14, Anastassov in view of Cho teach The method of claim 1, and further teach wherein the method is performed by at least one of:
a control system for an autonomous or semi-autonomous machine;
a perception system for an autonomous or semi-autonomous machine;
a system for performing simulation operations;
a system for performing real-time streaming;
a system for generating or presenting one or more of augmented reality content, virtual reality content, or mixed reality content;
a system for performing digital twin operations;
a system for performing deep learning operations;
a system implemented using an edge device;
a system implemented using a robot;
a system incorporating one or more virtual machines (VMs);
a system implemented at least partially in a data center;
a system for performing light transport simulation;
a system for performing collaborative content creation for 3D assets;
a system for generating synthetic data;
or a system implemented at least partially using cloud computing resources (Anastassov paragraph [0047] “the system 100 can process sensor data from one or more vehicles 103a-103n (also collectively referred to as vehicles 103) (e.g., standard vehicles, autonomous vehicles, heavily assisted driving (HAD) vehicles, semi-autonomous vehicles, etc.).”, paragraph [0119] “the machine learning system 125 can continuously provide and/or update a machine learning model (e.g., a support vector machine (SVM), neural network, decision tree, etc.) during training using, for instance, supervised deep convolution networks or equivalents.”).
Regarding claim 15, Anastassov teaches One or more processors comprising one or more circuits to (Anastassov paragraph [0162-0163] “FIG. 13 illustrates a computer system 1300 upon which an embodiment of the invention may be implemented. Computer system 1300 is programmed (e.g., via computer program code or instructions) to create a base map and identify special areas as described herein……One or more processors 1302 for processing information are coupled with the bus 1310.”): generate, for each cell of one or more cells of a grid representing a region of an environment (Anastassov paragraph [0038] “FIG. 2A is a diagram illustrating an example geo-spatial partitioning scheme, according to one embodiment. The diagram 200 shows an area of interest (e.g., a Washington D.C. map 201) divided into partitions 203 for the purposes of distributed/parallelized processing……By way of example, a partition 203m containing a portion of Washington D.C. borderline 201n is divided into grid cells 205”), a cell representation indicating one or more points that correspond with the cell and that represent corresponding sensor detections generated by one or more ego-machines in the environment (Anastassov teaches the probe data as the cell representation, paragraph [0008] “cause the apparatus to retrieve probe data collected from one or more sensors of one or more probe devices traveling within a geographic area including at least one geographic partition.” And paragraph [0036] “the system 100 can prepare probe data into seed points, then continue with the seed points (without the probe data)”); …… and generate, based at least on the lane data, a lane graph that represents the one or more lanes on one or more roads in the environment (Anastassov paragraph [0066] “the seed points 225 created in step 407 (e.g., FIGS. 2B-2C) can be used to create continuous road paths represented by polylines 1-6 in FIG. 2D.”, paragraph [0082] “In one embodiment, in step 411, the output module 309 can include the at least one continuous road path in an output representing a base map of the geographic area. In other embodiments, the base map further includes connections, intersections, splits/merges, etc. as a graph.”).
Anastassov does not explicitly teach apply the cell representations for the one or more decoders of a transformer machine learning model; generate, as output of the one or more decoders, lane data indicating one or more lanes associated with the one or more cells, one or more cross-sections of the one or more lanes associated with the one or more cells, and one or more connections between the one or more cross-sections……
Cho teaches apply the cell representations for the one or more decoders of a transformer machine learning model (Cho Figure 2, paragraph [0024] “With the output set of tensors, the computing system on the ego can apply a third set of encoders (e.g., an autoregressive decoder) to generate a set of tokens for the graph. Each token can define various properties of a point forming one or more of the lane segments through the environment.”, paragraph [0074-0075] “The tensor 250 can include the aggregate set of embeddings derived from both the sensor data and the map data …… The autoregressive decoder of the lane encoder 260 can include a sequence of models to process portions of input embeddings from the tensor 250 to generate output embeddings dependent on another output embedding derived from prior portion of input embeddings.” And paragraph [0086] “The point attribute predictor 316 can correspond to a portion of a machine learning (ML) model (e.g., the lane encoder 260 as discussed herein), and can include a set of weights arranged in accordance with a cross-attention, a self-attention, and a transformer”); generate, as output of the one or more decoders, lane data indicating one or more lanes associated with the one or more cells, one or more cross-sections of the one or more lanes associated with the one or more cells, and one or more connections between the one or more cross-sections…… (Cho teaches output tokens from machine learning model, further teaches tokens with point property and connectivity information. Cho paragraph [0063] “The output can include environment features (e.g., attributes gathered from the sensor data), map features (e.g., attributes in navigation map such as topological features and road layouts), classifications (e.g., a type of topology), and an output token (e.g., a combination of environment features, map features, and classifications) to be included in a graph defining lane segments”, paragraph [0075] “From processing the tensor 250, the lane encoder 260 can produce, output, or otherwise generate a set of lane instances 265 and at least one adjacency matrix 270. The lane instances 265 can define one or more lane segments through which the ego can potentially navigate the environment. The adjacent matrix 270 can define a connectivity or relationship among the lane segments defined in the lane segments 265.” And paragraph [0024] “Each token can define various properties of a point forming one or more of the lane segments through the environment ……. the computing system can classify the point by a type of topology, such as a starting point, a continuation, a fork, or a terminal point”);
Anastassov and Cho are in the same field of endeavor, namely computer graphics, especially in the field of road map generation based on sensor data. Cho teaches using an AI model with decoder to improve accuracy (Cho paragraph [0029] “the data ingested and/or predicted by the AI model(s) 110c with respect to the egos 140 (at inference time) may also be used to improve the AI model(s) 110c. Therefore, the system 100 depicts a continuous loop that can periodically improve the accuracy of the AI model(s) 110c. Moreover, the system 100 depicts a loop in which data received the egos 140 can be used to at training phase in addition to the inference phase.”). Therefore, it would have been obvious for a person of ordinary skill in the art before the effective filing date of the claimed invention to combine the teaching of Cho with the method of Anastassov to achieve accuracy and efficiency.
Regarding claim 17, Anastassov in view of Cho teach The one or more processors of claim 15, and further teach wherein the one or more processors are comprised in at least one of:
a control system for an autonomous or semi-autonomous machine;
a perception system for an autonomous or semi-autonomous machine;
a system for performing simulation operations;
a system for performing digital twin operations;
a system for performing light transport simulation;
a system for performing collaborative content creation for 3D assets;
a system for performing deep learning operations;
a system for performing remote operations;
a system for performing real-time streaming;
a system for generating or presenting one or more of augmented reality content, virtual reality content, or mixed reality content;
a system implemented using an edge device;
a system implemented using a robot;
a system for performing conversational AI operations;
a system implementing one or more language models;
a system implementing one or more large language models (LLMs);
a system for generating synthetic data;
a system for generating synthetic data using AI;
a system incorporating one or more virtual machines (VMs);
a system implemented at least partially in a data center;
or a system implemented at least partially using cloud computing resources (Anastassov paragraph [0047] “the system 100 can process sensor data from one or more vehicles 103a-103n (also collectively referred to as vehicles 103) (e.g., standard vehicles, autonomous vehicles, heavily assisted driving (HAD) vehicles, semi-autonomous vehicles, etc.).”, paragraph [0119] “the machine learning system 125 can continuously provide and/or update a machine learning model (e.g., a support vector machine (SVM), neural network, decision tree, etc.) during training using, for instance, supervised deep convolution networks or equivalents.”).
Regarding claim 18, Anastassov teaches A system comprising: one or more circuits to (Anastassov paragraph [0019] “FIG. 1 is a diagram of a system capable of creating a base map and identifying special areas, according to one embodiment.”):
apply at least one cell representation indicating one or more points corresponding with a cell of a region and representing sensor detections generated by one or more ego-machines in an environment (Anastassov teaches the probe data as the cell representation, paragraph [0008] “cause the apparatus to retrieve probe data collected from one or more sensors of one or more probe devices traveling within a geographic area including at least one geographic partition.” And paragraph [0036] “the system 100 can prepare probe data into seed points, then continue with the seed points (without the probe data)” and paragraph [0038] “FIG. 2A is a diagram illustrating an example geo-spatial partitioning scheme, according to one embodiment. The diagram 200 shows an area of interest (e.g., a Washington D.C. map 201) divided into partitions 203 for the purposes of distributed/parallelized processing……By way of example, a partition 203m containing a portion of Washington D.C. borderline 201n is divided into grid cells 205”) …… and generate, based at least on the lane data, a lane graph that represents the one or more lanes on one or more roads in the environment (Anastassov paragraph [0066] “the seed points 225 created in step 407 (e.g., FIGS. 2B-2C) can be used to create continuous road paths represented by polylines 1-6 in FIG. 2D.”, paragraph [0082] “In one embodiment, in step 411, the output module 309 can include the at least one continuous road path in an output representing a base map of the geographic area. In other embodiments, the base map further includes connections, intersections, splits/merges, etc. as a graph.”).
Anastassov is not relied on for the below claim language ……to one or more decoders of one or more neural networks; generate, as output of the one or more decoders, lane data indicating one or more lanes associated with the cell of the region, one or more cross-sections of the one or more lanes associated with the cell of the regions, and one or more connections between the one or more cross-sections ……
Cho teaches ……to one or more decoders of one or more neural networks (Cho Figure 2, paragraph [0024] “With the output set of tensors, the computing system on the ego can apply a third set of encoders (e.g., an autoregressive decoder) to generate a set of tokens for the graph. Each token can define various properties of a point forming one or more of the lane segments through the environment.”, paragraph [0074-0075] “The tensor 250 can include the aggregate set of embeddings derived from both the sensor data and the map data …… The autoregressive decoder of the lane encoder 260 can include a sequence of models to process portions of input embeddings from the tensor 250 to generate output embeddings dependent on another output embedding derived from prior portion of input embeddings.”); generate, as output of the one or more decoders, lane data indicating one or more lanes associated with the cell of the region, one or more cross-sections of the one or more lanes associated with the cell of the regions, and one or more connections between the one or more cross-sections …… (Cho teaches output tokens from machine learning model, further teaches tokens with point property and connectivity information. Cho paragraph [0063] “The output can include environment features (e.g., attributes gathered from the sensor data), map features (e.g., attributes in navigation map such as topological features and road layouts), classifications (e.g., a type of topology), and an output token (e.g., a combination of environment features, map features, and classifications) to be included in a graph defining lane segments”, paragraph [0075] “From processing the tensor 250, the lane encoder 260 can produce, output, or otherwise generate a set of lane instances 265 and at least one adjacency matrix 270. The lane instances 265 can define one or more lane segments through which the ego can potentially navigate the environment. The adjacent matrix 270 can define a connectivity or relationship among the lane segments defined in the lane segments 265.” And paragraph [0024] “Each token can define various properties of a point forming one or more of the lane segments through the environment ……. the computing system can classify the point by a type of topology, such as a starting point, a continuation, a fork, or a terminal point”);
Anastassov and Cho are in the same field of endeavor, namely computer graphics, especially in the field of road map generation based on sensor data. Cho teaches using an AI model with decoder to improve accuracy (Cho paragraph [0029] “the data ingested and/or predicted by the AI model(s) 110c with respect to the egos 140 (at inference time) may also be used to improve the AI model(s) 110c. Therefore, the system 100 depicts a continuous loop that can periodically improve the accuracy of the AI model(s) 110c. Moreover, the system 100 depicts a loop in which data received the egos 140 can be used to at training phase in addition to the inference phase.”). Therefore, it would have been obvious for a person of ordinary skill in the art before the effective filing date of the claimed invention to combine the teaching of Cho with the method of Anastassov to achieve accuracy and efficiency.
Regarding claim 20, Anastassov in view of Cho teach The system of claim 18, and further teach wherein the system is comprised in at least one of:
a control system for an autonomous or semi-autonomous machine;
a perception system for an autonomous or semi-autonomous machine;
a system for performing simulation operations;
a system for performing digital twin operations;
a system for performing light transport simulation;
a system for performing collaborative content creation for 3D assets;
a system for performing deep learning operations;
a system for performing real-time streaming;
a system for generating or presenting one or more of augmented reality content, virtual reality content, or mixed reality content;
a system implemented using an edge device;
a system implemented using a robot;
a system for performing conversational AI operations;
a system for generating synthetic data;
a system incorporating one or more virtual machines (VMs);
a system implemented at least partially in a data center;
or a system implemented at least partially using cloud computing resources (Anastassov paragraph [0047] “the system 100 can process sensor data from one or more vehicles 103a-103n (also collectively referred to as vehicles 103) (e.g., standard vehicles, autonomous vehicles, heavily assisted driving (HAD) vehicles, semi-autonomous vehicles, etc.).”, paragraph [0119] “the machine learning system 125 can continuously provide and/or update a machine learning model (e.g., a support vector machine (SVM), neural network, decision tree, etc.) during training using, for instance, supervised deep convolution networks or equivalents.”).
Claim(s) 2-4, 6 and 16 is/are rejected under 35 U.S.C. 103 as being unpatentable over Anastassov (US 20220161817 A1), hereinafter as Anastassov, in view of Cho et al. (US 20260109373 A1), hereinafter as Cho, further in view of Kaku et al. (US 20250146835 A1), hereinafter as Kaku.
Regarding claim 2, Anastassov in view of Cho teach The method of claim 1, but are not relied on for the below claim language wherein the one or more decoders comprise a cross- section decoder that outputs one or more indications of cross-sections of the one or more lanes associated with the one or more cells. Kaku teaches wherein the one or more decoders comprise a cross- section decoder that outputs one or more indications of cross-sections of the one or more lanes associated with the one or more cells (Kaku teaches using lateral slices as cell and further teaches a neural decoder based on the lateral slices to output lane boundaries indication, paragraph [0005] “the neural model includes a decoder that computes confidence values and boundary placements for the lane boundaries using a histogram of the aggregated features…...the estimation system generates a map with updated and fuller lane boundaries by processing sliced data individually and linking slices, thereby improving the accuracy and efficiency of generating maps”).
Anastassov, Cho and Kaku are in the same field of endeavor, namely computer graphics, especially in the field of road map generation based on sensor data. Kaku teaches using lateral slices with decoder to achieve accuracy and efficiency (Kaku paragraph [0017] “the estimation system improves the definition of lane boundaries and reduces computation costs by slicing road data that allows simpler geometric modeling and map generation.” And paragraph [0029] “slicing data has benefits because segments can solve a locally optimizable problem repeatedly. Slicing also improves solving disconnections or merging among inputs that impact decoding and improving global inferences as detection accuracy among lateral slices increases. For FIG. 3, the neural model 300 may select channels from lateral slices having keypoints that will expand detection areas and improve accuracy by applying various factoring.”). Therefore, it would have been obvious for a person of ordinary skill in the art before the effective filing date of the claimed invention to combine the teaching of Kaku with the method of Anastassov and Cho to achieve accuracy and efficiency.
Regarding claim 3, Anastassov in view of Cho and Kaku teach The method of claim 2, and further teach wherein generating the lane graph comprises aggregating at least two cross-sections associated with a bin in the region (Kaku paragraph [0016-0017] “the histogram can aggregate and compress features with reduced dimensions through bins that are each associated with a lateral slice, thereby improving efficiency……This can involve counting compressed data within the bins for relevancy and correlation of features. In one approach, the estimation system automatically recombines the lane boundaries individually along the road edges for generating a map, such as by merging lateral slices being adjacent that have defined features……the estimation system selects the features using the neural model by factoring a distance between compressed data in the bins and the relationship with the lane boundary.”).
Anastassov, Cho and Kaku are in the same field of endeavor, namely computer graphics, especially in the field of road map generation based on sensor data. Kaku teaches using lateral slices with decoder to achieve accuracy and efficiency (Kaku paragraph [0017] “the estimation system improves the definition of lane boundaries and reduces computation costs by slicing road data that allows simpler geometric modeling and map generation.” And paragraph [0029] “slicing data has benefits because segments can solve a locally optimizable problem repeatedly. Slicing also improves solving disconnections or merging among inputs that impact decoding and improving global inferences as detection accuracy among lateral slices increases. For FIG. 3, the neural model 300 may select channels from lateral slices having keypoints that will expand detection areas and improve accuracy by applying various factoring.”). Therefore, it would have been obvious for a person of ordinary skill in the art before the effective filing date of the claimed invention to combine the teaching of Kaku with the method of Anastassov and Cho to achieve accuracy and efficiency.
Regarding claim 4, Anastassov in view of Cho and Kaku teach The method of claim 2, and further teach wherein generating the lane graph comprises stitching at least two cross-sections together at least based on proximity or orientation of the at least two cross-sections relative to one another (Kaku paragraph [0016] “the estimation system automatically recombines the lane boundaries individually along the road edges for generating a map, such as by merging lateral slices being adjacent that have defined features.”).
Anastassov, Cho and Kaku are in the same field of endeavor, namely computer graphics, especially in the field of road map generation based on sensor data. Kaku teaches using lateral slices with decoder to achieve accuracy and efficiency (Kaku paragraph [0017] “the estimation system improves the definition of lane boundaries and reduces computation costs by slicing road data that allows simpler geometric modeling and map generation.” And paragraph [0029] “slicing data has benefits because segments can solve a locally optimizable problem repeatedly. Slicing also improves solving disconnections or merging among inputs that impact decoding and improving global inferences as detection accuracy among lateral slices increases. For FIG. 3, the neural model 300 may select channels from lateral slices having keypoints that will expand detection areas and improve accuracy by applying various factoring.”). Therefore, it would have been obvious for a person of ordinary skill in the art before the effective filing date of the claimed invention to combine the teaching of Kaku with the method of Anastassov and Cho to achieve accuracy and efficiency.
Regarding claim 6, Anastassov in view of Cho and Kaku teach The method of claim 2, and further teach wherein the one or more decoders comprise a connection decoder to connect at least a first portion of a lane and a second portion of the lane (Kaku paragraph [0032] “the estimation system 170 can link lane boundaries individually along road edges heuristically using the confidence values and the boundary positions outputted per lateral slice. This can include identifying relationships between lane characteristics that satisfy a threshold for an inverse distance and feature clarity along a road edge. For example, two end lateral slices have a dashed line with elevated confidence values with a middle lateral slice that is adjacent and includes missing paint. As such, the estimation system 170 can reliably merge the lateral slices together using a dashed line across three lateral slices if within the threshold for confidence and position.” And paragraph [0031] “although the neural model 300 illustrates implementing two decoders, the neural model 300 could implement a 2-depth layer for outputting the RB and LB values……Another process for 1-to-1 uses a 1d convolutional process for outputting confidence values and boundary placements for a RB and LB per lateral slice. Such boundary positions use an inverse distance between the RB/LB and inferred features that the neural model 300 assembles into a map.).
Anastassov, Cho and Kaku are in the same field of endeavor, namely computer graphics, especially in the field of road map generation based on sensor data. Kaku teaches using lateral slices with decoder to achieve accuracy and efficiency (Kaku paragraph [0017] “the estimation system improves the definition of lane boundaries and reduces computation costs by slicing road data that allows simpler geometric modeling and map generation.” And paragraph [0029] “slicing data has benefits because segments can solve a locally optimizable problem repeatedly. Slicing also improves solving disconnections or merging among inputs that impact decoding and improving global inferences as detection accuracy among lateral slices increases. For FIG. 3, the neural model 300 may select channels from lateral slices having keypoints that will expand detection areas and improve accuracy by applying various factoring.”). Therefore, it would have been obvious for a person of ordinary skill in the art before the effective filing date of the claimed invention to combine the teaching of Kaku with the method of Anastassov and Cho to achieve accuracy and efficiency.
Regarding claim 16, claim 16 has similar limitations as claim 2, therefore it is rejected under the same rationale as claim 2.
Claim(s) 7-9 and 19 is/are rejected under 35 U.S.C. 103 as being unpatentable over Anastassov (US 20220161817 A1), hereinafter as Anastassov, in view of Cho et al. (US 20260109373 A1), hereinafter as Cho, further in view of NPL Liu et al. ("VectorMapNet: End-to-end Vectorized HD Map Learning"),hereinafter as Liu.
Regarding claim 7, Anastassov in view of Cho teach The method of claim 1, but are not relied on for the below claim language wherein the one or more decoders comprise an edge decoder that outputs one or more indications of Bezier curves or polyline parameterizations associated with edges of the one or more lanes. Liu teaches wherein the one or more decoders comprise an edge decoder that outputs one or more indications of Bezier curves or polyline parameterizations associated with edges of the one or more lanes (Liu teaches an edge decoder in the polyline generator. Page 5, right column, third paragraph, “the polyline generator focuses on the detailed geometry of HD map, which entails calculating variable-length polyline vertices and their order.” And Page 6, right column, first paragraph, “Each polyline’s keypoint coordinates and class label are tokenized and fed in as the query inputs of the transformer decoder. Then a sequence of vertex tokens are fed into the transformer iteratively, integrating BEV features with cross-attention, and decoded as polyline vertices.”).
Anastassov, Cho and Liu are in the same field of endeavor, namely computer graphics, especially in the field of road map generation based on sensor data. Liu teaches using VectorMapNet to generate the HD semantic map in order to achieve accuracy and efficiency (Liu Page 9, Right Column, fourth paragraph, “Our experiments show that VectorMapNet can generate coherent and complex geometries for urban map elements, benefiting from the polyline primitives. We believe that this novel way to learn HD maps provides a new perspective on the HD semantic map learning problem.”). Therefore, it would have been obvious for a person of ordinary skill in the art before the effective filing date of the claimed invention to combine the teaching of Liu with the method of Anastassov and Cho to achieve accuracy and efficiency.
Regarding claim 8, Anastassov in view of Cho and Liu teach The method of claim 7, and further teach wherein generating the lane graph includes projecting the one or more indications of the Bezier curves or the polyline parameterizations to a map view of the environment (Anastassov teaches a polyline of route 1117 on a map view of the environment. Figure 11, paragraph [0122] “a user interface (UI) 1100 (e.g., a navigation application 113) is generated for a UE 111 (e.g., a mobile device, an embedded navigation system, a client terminal, etc.) that includes a map 1101, …… However, the system 100 determines an optimum route 1109 which nevertheless involves a mobile obstacle 1111 (e.g., a street cleaning vehicle) and a work area 1113, and shows an alert: “Warning! Work Areas Detected Along Route.” In response to an input 1115 of “Show Alternative Route,” the UI 1100 presents an alternative route 1117.”).
Regarding claim 9, Anastassov in view of Cho and Liu teach The method of claim 8, and further teach wherein generating the lane graph further includes stitching the one or more indications of the Bezier curves or the polyline parameterizations together across neighboring regions in the map view of the environment (Anastassov paragraph [0081] “The most common connection type in across-tiles merging is a segment A of Tile 1 terminated by a node of valence 1 can be connected to segment B of Tile 2 starting at a node of valence 1 in FIG. 7A. The arrows indicate the direction of travel. In this case, a combined polyline is created in FIG. 7B, defining a new map segment A+B associated with Tile 2”).
Regarding claim 19, Anastassov in view of Cho teach The system of claim 18, but are not relied on for the below claim language wherein the one or more decoders comprise a cross- section decoder, a connection decoder, an edge decoder, or a combination thereof. Liu teaches wherein the one or more decoders comprise a cross- section decoder, a connection decoder, an edge decoder, or a combination thereof (Liu teaches an edge decoder in the polyline generator. Page 5, right column, third paragraph, “the polyline generator focuses on the detailed geometry of HD map, which entails calculating variable-length polyline vertices and their order.” And Page 6, right column, first paragraph, “Each polyline’s keypoint coordinates and class label are tokenized and fed in as the query inputs of the transformer decoder. Then a sequence of vertex tokens are fed into the transformer iteratively, integrating BEV features with cross-attention, and decoded as polyline vertices.”).
Anastassov, Cho and Liu are in the same field of endeavor, namely computer graphics, especially in the field of road map generation based on sensor data. Liu teaches using VectorMapNet to generate the HD semantic map in order to achieve accuracy and efficiency (Liu Page 9, Right Column, fourth paragraph, “Our experiments show that VectorMapNet can generate coherent and complex geometries for urban map elements, benefiting from the polyline primitives. We believe that this novel way to learn HD maps provides a new perspective on the HD semantic map learning problem.”). Therefore, it would have been obvious for a person of ordinary skill in the art before the effective filing date of the claimed invention to combine the teaching of Liu with the method of Anastassov and Cho to achieve accuracy and efficiency.
Claim(s) 5 is/are rejected under 35 U.S.C. 103 as being unpatentable over Anastassov (US 20220161817 A1), hereinafter as Anastassov, in view of Cho et al. (US 20260109373 A1), hereinafter as Cho, further in view of Kaku et al. (US 20250146835 A1), hereinafter as Kaku, and Chen et al. (US 20240067207 A1), hereinafter as Chen.
Regarding claim 5, Anastassov in view of Cho and Kaku teach The method of claim 2, but are not relied on for the below claim language further comprising using one or more trajectories to connect at least a first portion of a lane and a second portion of the lane. Chen teaches further comprising using one or more trajectories to connect at least a first portion of a lane and a second portion of the lane (Chen teaches using historical trajectory data in lane boundary detection of neural network, it would have been obvious for a person of ordinary skill in the art before the effective filing date of the claimed invention to combine the teaching of Chen with the decoder of Kaku, paragraph [0026] “lane-boundary detection training system 200 can communicate with other computing systems and local or remote databases to acquire image data 230 and historical vehicle trajectory data 235 for use in training a neural-network-based model to detect roadway lane boundaries.”).
Anastassov, Cho, Kaku and Chen are in the same field of endeavor, namely computer graphics, especially in the field of road map generation based on sensor data. Chen teaches using historical vehicle trajectory data to improve performance of detecting lane boundaries based on sensor data (Chen paragraph [0013] “First, during the training of a neural network to detect lane boundaries based on image input, historical vehicle trajectory data is used as weak supervision to improve the performance of the resulting trained network. Second, the trained network, when deployed in a semi-autonomous or autonomous vehicle (inference or test time), can continue to use historical vehicle trajectory data as an input to improve performance, particularly when HD map data is unavailable or when weather conditions (e.g., heavy rain or snow) interfere with the ability of the system to detect lane boundaries based on sensor (image) data.”). Therefore, it would have been obvious for a person of ordinary skill in the art before the effective filing date of the claimed invention to combine the teaching of Chen with the method of Anastassov, Cho and Kaku to achieve accuracy and efficiency.
Claim(s) 10 and 11 is/are rejected under 35 U.S.C. 103 as being unpatentable over Anastassov (US 20220161817 A1), hereinafter as Anastassov, in view of Cho et al. (US 20260109373 A1), hereinafter as Cho, further in view of NPL Liu et al. (“VectorMapNet: End-to-end Vectorized HD Map Learning”), hereinafter as Liu, and Pham et al. (US 20200341466 A1), hereinafter as Pham.
Regarding claim 10, Anastassov in view of Cho teach The method of claim 1, but are not relied on for the below claim language wherein the one or more decoders comprise an edge decoder that autoregressively outputs one or more indications of keypoints along one or more centerlines of the one or more lanes and one or more indications of offsets to lane edges of the one or more lanes. Liu teaches wherein the one or more decoders comprise an edge decoder that autoregressively outputs one or more indications of keypoints along one or more centerlines of the one or more lanes (Liu Page 6, Left Column, Last Paragraph, “Architecture. To model these local geometric structures of polylines, the autoregressive network we choose is Transformer (Vaswani et al., 2017) (see the bottom-right of Figure 2).”, Page 4, right column, first paragraph, “we divide the task into three distinct components: (1) A BEV feature extractor (§ 3.2) that lifts various sensor modality inputs into a canonical feature space. (2) A map element detector (§ 3.3) that locates and classifies all map elements by predicting element keypoints A = {Ai ∈ R k×2 |i = 1, . . . , N} and their class labels L = {li ∈ Z|i = 1, . . . , N}. The definition of element keypoint representation A is described in § 3.3. (3) A polyline generator (§ 3.4) that produces a sequence of ordered polyline vertices which describes the local geometry of each detected map element (Ai , li).”, and Page 8, Right Column, First paragraph, “Centerline prediction by VectorMapNet. As discussed in § 3.1 and above, the polyline is a versatile primitive, capable of representing map element classes that extend beyond the elements in the HD semantic map setting. To further demonstrate this flexibility, we expand VectorMapNet to predict the centerline, an imaginary line commonly used as a reference for driving direction, vehicle positioning, and navigation.”).
Anastassov, Cho and Liu are in the same field of endeavor, namely computer graphics, especially in the field of road map generation based on sensor data. Liu teaches using VectorMapNet to generate the HD semantic map in order to achieve accuracy and efficiency (Liu Page 9, Right Column, fourth paragraph, “Our experiments show that VectorMapNet can generate coherent and complex geometries for urban map elements, benefiting from the polyline primitives. We believe that this novel way to learn HD maps provides a new perspective on the HD semantic map learning problem.”). Therefore, it would have been obvious for a person of ordinary skill in the art before the effective filing date of the claimed invention to combine the teaching of Liu with the method of Anastassov and Cho to achieve accuracy and efficiency.
Anastassov in view of Cho and Liu fail to teach and one or more indications of offsets to lane edges of the one or more lanes. Pham teaches and one or more indications of offsets to lane edges of the one or more lanes (Pham teaches an decoder outputting the width of a lane corresponding to the key points as the offset to the lane edges, paragraph [0026] “The sensor data may be applied to a neural network (e.g., a deep neural network (DNN), such as a convolutional neural network (CNN)) that is trained to identify areas of interest pertaining to road markings, road boundaries, intersections……More specifically, the neural network may be designed to compute key points corresponding to segments of an intersection (e.g., corresponding to lanes, bike paths, etc., and/or corresponding to features therein—such as cross walks, intersection entry points, intersection exit points, etc.), and to generate outputs identifying, for each key point, a width of a lane corresponding to the key point, a directionality of the lane”).
Anastassov, Cho, Liu and Pham are in the same field of endeavor, namely computer graphics, especially in the field of road map generation based on sensor data. Pham teaches using a machine learning network to decide lanes and crosswalks in order to achieve accuracy and efficiency for road map (Pham paragraph [0007] “As a result of using live perception to generate an understanding of each intersection, the process of generating paths for navigating the intersection may be comparatively less time-consuming, less computationally intense, and more scalable as the system may learn to diagnose each intersection in real-time or near real-time, without requiring prior experience or knowledge of the intersection.”). Therefore, it would have been obvious for a person of ordinary skill in the art before the effective filing date of the claimed invention to combine the teaching of Pham with the method of Anastassov in view of Cho and Liu to achieve accuracy and efficiency of road map.
Regarding claim 11, Anastassov in view of Cho, Liu and Pham teach The method of claim 10, and further teach wherein generating the lane graph includes projecting the one or more indications of the keypoints and the one or more indications of the offsets to a map view of the environment (Liu teaches showing a map of the lane graph with the green boundary line and gray centerline. Each vector represents the direction of key points. Page 7, Right Column, Figure 5, “Figure 5: The centerline predictions by VectorMapNet, where the gray lines are the predicted centerlines.” And Page 5, Figure 3, “The arrow line indicates the direction of the example polyline, and the arrow dash lines indicate the vertices order of keypoint representations.”).
Anastassov, Cho, Liu and Pham are in the same field of endeavor, namely computer graphics, especially in the field of road map generation based on sensor data. Liu teaches using VectorMapNet to generate the HD semantic map in order to achieve accuracy and efficiency (Liu Page 9, Right Column, fourth paragraph, “Our experiments show that VectorMapNet can generate coherent and complex geometries for urban map elements, benefiting from the polyline primitives. We believe that this novel way to learn HD maps provides a new perspective on the HD semantic map learning problem.”). Therefore, it would have been obvious for a person of ordinary skill in the art before the effective filing date of the claimed invention to combine the teaching of Liu with the method of Anastassov, Cho and Pham to achieve accuracy and efficiency.
Allowable Subject Matter
Claim 12 objected to as being dependent upon a rejected base claim, but would be allowable if rewritten in independent form including all of the limitations of the base claim and any intervening claims.
The following is a statement of reasons for the indication of allowable subject matter:
Regarding claim 12, the closest prior art of Pham teaches deciding center key points, left and right edge key points, further teaches connecting pair of key points together to generate a path (paragraph [0078] “The heat map(s) 108 may be used by the decoder to determine locations of the center key points, left key points, and/or right key points corresponding to a center of a lane, a left edge of a lane, or a right edge of a lane, respectively.” And paragraph [0025] “Computer vision and/or machine learning algorithm(s) may be trained to detect the key points of an intersection, and the (center) key points may be connected together—using one or more filters—to generate paths and/or trajectories for the vehicle to effectively and accurately navigate the intersection.”). However, Pham fails to teach the combined limitation as a whole, “wherein generating the lane graph further includes stitching the one or more indications of the keypoints and the one or more indications of the offsets across neighboring regions in the map view of the environment”. Furthermore, no prior art of record either alone or in combination teaches the above limitation as a whole. Therefore, claim 12 is considered to allowable.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Walls et al. (US 20250095293 A1) teaches a machine learning method to decode features from sliced data to form lane structure of road graph.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to XIAOMING WEI whose telephone number is (571)272-3831. The examiner can normally be reached M-F 8:00-5:00.
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, Kee Tung can be reached at (571)272-7794. 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.
/KEE M TUNG/Supervisory Patent Examiner, Art Unit 2611
/XIAOMING WEI/ Examiner, Art Unit 2611
Prosecution Insights