Search

JP-7856489-B2 - Neural network routing planning

JP7856489B2JP 7856489 B2JP7856489 B2JP 7856489B2JP-7856489-B2

Inventors

  • シュエリン リー
  • シフェイ リュー
  • シャリー二 デ メッロ
  • ヤン カウツ

Assignees

  • エヌビディア コーポレーション

Dates

Publication Date
20260511
Application Date
20220601
Priority Date
20210615

Claims (20)

  1. It is a processor, Equipped with one or more circuits, The one or more circuits use one or more neural networks to calculate multiple paths from different starting positions to one or more destination positions that the autonomous device is to traverse, at least in part on the outputs of one or more of the one or more neural networks, which include the distances from the different starting positions to the one or more destination positions . The aforementioned one or more neural networks include an implicit environment function that represents the environment via an environment field. The distance is the distance to be reached, which is obtained as the output of the implicit environment function, which has been trained to represent the environment field as a mapping from location coordinates to an arrival time function u(x). The aforementioned arrival time function u(x) is a processor obtained by solving a continuous shortest path problem that represents the aforementioned arrival time function u(x) .
  2. The aforementioned one or more circuits shall be at least, To obtain the first location, the set of locations, and the final location, The one or more neural networks are made to calculate a set of distances based at least partially on the set of locations and the final location. The processor according to claim 1, which calculates the plurality of paths based at least in part on the set of distances, wherein the plurality of paths form a path from the first location to the final location, and uses the one or more neural networks to calculate the plurality of paths.
  3. The first location described above is the location of the autonomous device, A subset of locations from the set of locations is accessible to the autonomous device from the first location. The processor according to claim 2.
  4. The aforementioned one or more circuits include at least, Obtaining a subset of distances from the set of distances corresponding to the subset of locations, Selecting a second location from the subset of locations based at least partially on the subset of distances, The processor according to claim 3, which calculates the first path by calculating the first path among a plurality of paths, including the path from the first location to the second location.
  5. The processor according to claim 4, wherein the second location corresponds to the minimum distance among the subset of distances.
  6. The processor according to claim 2, wherein one or more neural networks calculate the set of distances in a single forward pass.
  7. The processor according to claim 2, wherein the distances in the set of distances correspond to the distances along the path from one of the locations in the set of locations to the final location.
  8. A machine-readable medium, The machine-readable medium stores a set of instructions. If the set of instructions is executed by one or more processors, the one or more processors will use one or more neural networks to calculate multiple paths from different starting positions to one or more destination positions that the autonomous device is to traverse, based at least in part on one or more outputs of the one or more neural networks, which include the distances from the different starting positions to the one or more destination positions . The aforementioned one or more neural networks include an implicit environment function that represents the environment via an environment field. The distance is the distance to be reached, which is obtained as the output of the implicit environment function, which has been trained to represent the environment field as a mapping from location coordinates to an arrival time function u(x). The aforementioned arrival time function u(x) is a machine-readable medium obtained by solving a continuous shortest path problem that represents the aforementioned arrival time function u(x) .
  9. If the set of instructions is executed by one or more processors, the one or more processors shall: To acquire the characteristics of the environment, Selecting a location in the aforementioned environment, The machine-readable medium according to claim 8, further comprising instructions for inputting at least the features and the locations into the one or more neural networks in order to obtain a plurality of distances corresponding to a plurality of locations in the environment.
  10. If the set of instructions is executed by the one or more processors, the one or more processors will: Selecting a set of locations accessible to the autonomous device, Obtaining a set of distances from among the multiple distances corresponding to the set of locations, The machine-readable medium according to claim 9, further comprising an instruction to cause the machine-readable medium to select a first location from the set of locations based at least in part on the set of distances, wherein the first path from the plurality of paths indicates a path from the autonomous device to the first location.
  11. The machine-readable medium according to claim 10, wherein the set of instructions, when executed by the one or more processors, further comprises instructions causing the one or more processors to cause the autonomous device to navigate to the first location using the first path.
  12. The machine-readable medium according to claim 11, wherein the autonomous device is an autonomous vehicle.
  13. The machine-readable medium according to claim 9, wherein the aforementioned features are generated by one or more encoders based on the representation of the environment.
  14. The machine-readable medium according to claim 13, wherein the representation of the environment is an image or a point cloud.
  15. It is a system, A computer comprising one or more computers having one or more processors, The one or more processors use one or more neural networks to calculate multiple paths from different starting positions to one or more destination positions that the autonomous device is to traverse, at least in part on the outputs of the one or more neural networks, which include the distances from the different starting positions to the one or more destination positions . The aforementioned one or more neural networks include an implicit environment function that represents the environment via an environment field. The distance is the distance to be reached, which is obtained as the output of the implicit environment function, which has been trained to represent the environment field as a mapping from location coordinates to an arrival time function u(x). The aforementioned arrival time function u(x) is a system obtained by solving a continuous shortest path problem that represents the aforementioned arrival time function u(x) .
  16. The aforementioned one or more processors further, Capturing the representation of the environment, The system according to claim 15, for using one or more neural networks to calculate the plurality of paths from a first location to a second location of the autonomous device in the environment.
  17. The system according to claim 16, wherein the one or more processors are further configured to use the one or more neural networks to compute one or more distance values for one or more locations in the environment, based at least in part on the representation of the environment.
  18. The aforementioned one or more processors further, Calculating the size of the step of the autonomous device, Selecting a set of locations accessible from the first location of the autonomous device through the steps, The system according to claim 17, which is for selecting a third location from the set of locations based at least partially on one or more distance values.
  19. The system according to claim 16, wherein the representation of the environment is captured through one or more depth cameras.
  20. The system according to claim 16, wherein the representation of the environment is a 2D representation or a 3D representation.

Description

At least one embodiment relates to a processing resource used to compute a path through an environment using a neural network. For example, at least one embodiment relates to a processor or computing resource used to compute the distance to a location in an environment using a neural network to compute a path according to various novel techniques described herein. Calculating paths through an environment is a crucial task in many contexts. In various cases, calculating paths to a target through an environment can be difficult, such as when the target and environment undergo various changes. Calculating paths to a target through an environment can also require significant computing resources. Therefore, techniques for calculating paths through an environment can be improved. "Taxonomy and Definitions for Terms Related to Driving Automation Systems for On-Road Motor Vehicles" (for example, standard No. J3016-201806, issued June 15, 2018; standard No. J3016-201609, issued September 30, 2016; and previous and newer versions of this standard) This figure shows an example of an implicit environment function for an environment, based on at least one embodiment.This figure shows one example of an implicit environment function for different target locations in the environment, based on at least one embodiment.This figure shows one example of an implicit environment function for different target locations and environments, based on at least one embodiment.This figure shows an example of an agent that uses an implicit environment function to navigate to a target, based on at least one embodiment.This figure shows an example of an environment field for a multi-user navigation environment, based on at least one embodiment.This figure shows one example of the use of a negative environment field for a 3D environment, according to at least one embodiment.This figure shows an example of the results obtained using an implicit environment function, based on at least one embodiment.This figure shows an example of the results of agent navigation in at least one embodiment.This figure shows an example of the results obtained by fitting a given human sequence to a searched trajectory in a 3D indoor environment, according to at least one embodiment.This figure shows an example of the results obtained by fitting a given human sequence to a searched trajectory in a 3D indoor environment, according to at least one embodiment.This figure shows an example of a process for calculating multiple paths using an implicit environment function, based on at least one embodiment.This figure shows the inference and/or training logic according to at least one embodiment.This figure shows the inference and/or training logic according to at least one embodiment.This figure shows the training and deployment of a neural network in at least one embodiment.This figure shows an exemplary data center system according to at least one embodiment.This figure shows an example of an autonomous vehicle based on at least one embodiment.This figure shows an example of camera location and field of view for the autonomous vehicle shown in Figure 15A, according to at least one embodiment.This block diagram shows an exemplary system architecture for an autonomous vehicle, according to at least one embodiment, as shown in Figure 15A.This figure shows a system for communication between one or more cloud-based servers and the autonomous vehicle shown in Figure 15A, according to at least one embodiment.A block diagram of a computer system according to at least one embodiment.A block diagram of a computer system according to at least one embodiment.This figure shows a computer system according to at least one embodiment.This figure shows a computer system according to at least one embodiment.This figure shows a computer system according to at least one embodiment.This figure shows a computer system according to at least one embodiment.This figure shows a computer system according to at least one embodiment.This figure shows a computer system according to at least one embodiment.This figure shows a shared programming model based on at least one embodiment.This figure shows a shared programming model based on at least one embodiment.This figure shows an exemplary integrated circuit and associated graphics processor according to at least one embodiment.This figure shows an exemplary integrated circuit and associated graphics processor according to at least one embodiment.This figure shows an exemplary integrated circuit and associated graphics processor according to at least one embodiment.This figure shows additional exemplary graphics processor logic in at least one embodiment.This figure shows additional exemplary graphics processor logic in at least one embodiment.This figure shows a computer system according to at least one embodiment.This figure shows a parallel processor according to at least one embodiment.This figure shows a partition unit according to at least one