Abstract: The described aspects and implementations include a method for using transformers to generate maps for use by AVs. The method includes generating an input embedding based, at least in part, on sensing data from a sensing system of the AV; selecting one or more transformer decoder queries directing a transformer decoder to a particular portion of the input embedding; generating, using the one or more transformer decoder queries and the input embedding as input to the transformer decoder, one or more driving environment embeddings for a navigation system of the AV, and each driving environment embedding comprises a vector representation of a feature of the driving environment; providing the one or more driving environment embeddings to the navigation system of the AV, wherein the navigation system is configured to navigate the AV in the driving environment based, at least in part, on the one or more driving environment embeddings.

Abstract: A system includes a memory device, and a processing device, operatively coupled to the memory device, to receive a set of input data including a roadgraph and a distribution of a plurality of scene types associated with a driving context, the roadgraph including an autonomous vehicle driving path, perform stochastic sampling based on the distribution of the plurality of scene types to obtain a set of synthetic scenes for the driving context, and train a machine learning model to identify, for each synthetic scene of the set of synthetic scenes, a respective modified autonomous vehicle driving path.

Abstract: The technology relates to approaches for determining appropriate stopping locations at intersections for vehicles operating in a self-driving mode. While many intersections have stop lines painted on the roadway, many others have no such lines. Even if a stop line is present, the physical location may not match what is in store map data, which may be out of date due to construction or line repainting. Aspects of the technology employ a neural network that utilizes input training data and detected sensor data to perform classification, localization and uncertain estimation processes. Based on these processes, the system is able to evaluate distribution information for possible stop locations. The vehicle uses such information to determine an optimal stop point, which may or may not correspond to a stop line in the map data. This information is also used to update the existing map data, which can be shared with other vehicles.

Abstract: A system identifies a set of input data including a roadgraph identifying an intermediate autonomous vehicle (AV) driving path related to a scene representing an environment proximate an AV. The intermediate AV driving path reflects a modification to an initial (AV) driving path to avoid one or more obstacles obstructing the initial AV driving path. The system performs, using the set of input data, a path adjustment operation that identifies one or more candidate AV driving paths based on the intermediate AV driving path and determines a cost value for each of the candidate AV driving paths. The system identifies, among the candidate AV driving paths, a final AV driving path having a cost value that satisfies an evaluation criterion. The final AV driving path is to be included in the set of training data as a target output paired with training input including scene data identifying the scene.

Abstract: The technology relates to approaches for determining appropriate stopping locations at intersections for vehicles operating in a self-driving mode. While many intersections have stop lines painted on the roadway, many others have no such lines. Even if a stop line is present, the physical location may not match what is in store map data, which may be out of date due to construction or line repainting. Aspects of the technology employ a neural network that utilizes input training data and detected sensor data to perform classification, localization and uncertain estimation processes. Based on these processes, the system is able to evaluate distribution information for possible stop locations. The vehicle uses such information to determine an optimal stop point, which may or may not correspond to a stop line in the map data. This information is also used to update the existing map data, which can be shared with other vehicles.

Abstract: The technology relates to approaches for determining appropriate stopping locations at intersections for vehicles operating in a self-driving mode. While many intersections have stop lines painted on the roadway, many others have no such lines. Even if a stop line is present, the physical location may not match what is in store map data, which may be out of date due to construction or line repainting. Aspects of the technology employ a neural network that utilizes input training data and detected sensor data to perform classification, localization and uncertain estimation processes. Based on these processes, the system is able to evaluate distribution information for possible stop locations. The vehicle uses such information to determine an optimal stop point, which may or may not correspond to a stop line in the map data. This information is also used to update the existing map data, which can be shared with other vehicles.

Abstract: A system identifies a set of input data including a roadgraph identifying an intermediate autonomous vehicle (AV) driving path related to a scene representing an environment proximate an AV. The intermediate AV driving path reflects a modification to an initial (AV) driving path to avoid one or more obstacles obstructing the initial AV driving path. The system performs, using the set of input data, a path adjustment operation that identifies one or more candidate AV driving paths based on the intermediate AV driving path and determines a cost value for each of the candidate AV driving paths. The system identifies, among the candidate AV driving paths, a final AV driving path having a cost value that satisfies an evaluation criterion. The final AV driving path is to be included in the set of training data as a target output paired with training input including scene data identifying the scene.

Abstract: The technology relates to approaches for determining appropriate stopping locations at intersections for vehicles operating in a self-driving mode. While many intersections have stop lines painted on the roadway, many others have no such lines. Even if a stop line is present, the physical location may not match what is in store map data, which may be out of date due to construction or line repainting. Aspects of the technology employ a neural network that utilizes input training data and detected sensor data to perform classification, localization and uncertain estimation processes. Based on these processes, the system is able to evaluate distribution information for possible stop locations. The vehicle uses such information to determine an optimal stop point, which may or may not correspond to a stop line in the map data. This information is also used to update the existing map data, which can be shared with other vehicles.

Abstract: The technology relates to approaches for determining appropriate stopping locations at intersections for vehicles operating in a self-driving mode. While many intersections have stop lines painted on the roadway, many others have no such lines. Even if a stop line is present, the physical location may not match what is in store map data, which may be out of date due to construction or line repainting. Aspects of the technology employ a neural network that utilizes input training data and detected sensor data to perform classification, localization and uncertain estimation processes. Based on these processes, the system is able to evaluate distribution information for possible stop locations. The vehicle uses such information to determine an optimal stop point, which may or may not correspond to a stop line in the map data. This information is also used to update the existing map data, which can be shared with other vehicles.

Abstract: A system includes a memory device, and a processing device, operatively coupled to the memory device, to receive a set of input data including a roadgraph, the roadgraph including an autonomous vehicle driving path, modify the roadgraph to obtain a modified roadgraph by adjusting a trajectory of the autonomous vehicle driving path, place a set of artifacts along one or more lane boundaries of the modified roadgraph to generate a synthetic scene, and train a machine learning model used to navigate an autonomous vehicle based on the synthetic scene.

Abstract: A system includes a memory device, and a processing device, operatively coupled to the memory device, to receive a set of input data including a roadgraph. The roadgraph includes an autonomous vehicle driving path. The processing device is further to determine that the autonomous vehicle driving path is affected by one or more obstacles, identify a set of candidate paths that avoid the one or more obstacles, each candidate path of the set of candidate paths being associated with a cost value, select, from the set of candidate paths, a candidate path with an optimal cost value to obtain a selected candidate path, generate a synthetic scene based on the selected candidate path, and train a machine learning model to navigate an autonomous vehicle based on the synthetic scene.

Abstract: The technology relates to approaches for determining appropriate stopping locations at intersections for vehicles operating in a self-driving mode. While many intersections have stop lines painted on the roadway, many others have no such lines. Even if a stop line is present, the physical location may not match what is in store map data, which may be out of date due to construction or line repainting. Aspects of the technology employ a neural network that utilizes input training data and detected sensor data to perform classification, localization and uncertain estimation processes. Based on these processes, the system is able to evaluate distribution information for possible stop locations. The vehicle uses such information to determine an optimal stop point, which may or may not correspond to a stop line in the map data. This information is also used to update the existing map data, which can be shared with other vehicles.