IP Library › Granted Patent US 12,450,469
Granted Patent B1
US 12,450,469 · App. 17/842,605 · Granted Oct 21, 2025

Anomaly estimation neural networks for behavior prediction

Inventor: Khaled Refaat (Mountain View, CA)
Assignee: Waymo LLC
G06N3/045B60W60/0027B60W2556/45
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12,450,469
App. No.
17/842,605
Granted
Oct 21, 2025
Kind
B1
Abstract

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for anomaly estimation for behavior predictions using a neural network. One of the methods includes receiving data characterizing a scene that includes an agent in an environment. A behavior prediction input generated from the data is processed using a behavior prediction model. The behavior prediction model is configured to process the behavior prediction input to generate a predicted probability distribution over a plurality of possible behaviors for the agent. An anomaly estimation input generated from the data is processed using an anomaly estimation model. The anomaly estimation model is configured to process the anomaly estimation input to generate a prediction error for the predicted probability distribution. The prediction error indicates an error between the predicted probability distribution generated by the behavior prediction model and another predicted probability distribution generated by another behavior prediction model.

Claims (38)

1. A method comprising:

receiving data characterizing a scene that includes an agent in an environment;

processing a behavior prediction input generated from the data using a behavior prediction neural network, wherein the behavior prediction neural network is configured to receive the behavior prediction input and to process the behavior prediction input to generate a predicted probability distribution over a plurality of possible behaviors for the agent; and

processing an anomaly estimation input generated from the data characterizing the scene that includes the agent using an anomaly estimation neural network, wherein the anomaly estimation neural network is configured to receive the anomaly estimation input and to process the anomaly estimation input to generate a prediction error for the predicted probability distribution, wherein the prediction error indicates an error between the predicted probability distribution over the plurality of possible behaviors generated by the behavior prediction neural network and another predicted probability distribution over the plurality of possible behaviors that would be generated by another behavior prediction neural network from the same behavior prediction input.

2. The method of claim 1 , wherein the behavior prediction neural network is an on-board behavior prediction neural network that runs on-board an autonomous vehicle, and wherein the other behavior prediction neural network is an off-board behavior prediction neural network that consumes more resources than the on-board behavior prediction neural network.

3. The method of claim 2 , wherein the off-board behavior prediction neural network consumes more computational resources than a compute budget allocated for behavior prediction on-board the autonomous vehicle.

4. The method of claim 1 , wherein the behavior prediction neural network has fewer parameters than the other behavior prediction neural network.

5. The method of claim 1 , wherein the prediction error is smaller than a threshold value, further comprising:

generating a planned trajectory for an autonomous vehicle based on the predicted probability distribution.

6. The method of claim 1 , wherein the anomaly estimation input is generated from the data characterizing the scene that includes the agent and the predicted probability distribution.

7. The method of claim 1 , wherein the prediction error is not smaller than the threshold value, further comprising:

generating a planned trajectory for an autonomous vehicle based on the predicted probability distribution; and

modifying a planned trajectory for an autonomous vehicle to account for the predicted probability distribution possibly being erroneous.

8. A system comprising one or more computers and one or more storage devices storing instructions that when executed by the one or more computers cause the one or more computers to perform operations comprising:

receiving data characterizing a scene that includes an agent in an environment;

processing a behavior prediction input generated from the data using a behavior prediction neural network, wherein the behavior prediction neural network is configured to receive the behavior prediction input and to process the behavior prediction input to generate a predicted probability distribution over a plurality of possible behaviors for the agent; and

processing an anomaly estimation input generated from the data characterizing the scene that includes the agent using an anomaly estimation neural network, wherein the anomaly estimation neural network is configured to receive the anomaly estimation input and to process the anomaly estimation input to generate a prediction error for the predicted probability distribution, wherein the prediction error indicates an error between the predicted probability distribution over the plurality of possible behaviors generated by the behavior prediction neural network and another predicted probability distribution over the plurality of possible behaviors that would be generated by another behavior prediction neural network from the same behavior prediction input.

9. The system of claim 8 , wherein the behavior prediction neural network is an on-board behavior prediction neural network that runs on-board an autonomous vehicle, and wherein the other behavior prediction neural network is an off-board behavior prediction neural network that consumes more resources than the on-board behavior prediction neural network.

10. The system of claim 9 , wherein the off-board behavior prediction neural network consumes more computational resources than a compute budget allocated for behavior prediction on-board the autonomous vehicle.

11. The system of claim 8 , wherein the behavior prediction neural network has fewer parameters than the other behavior prediction neural network.

12. The system of claim 8 , wherein the prediction error is smaller than a threshold value, the operations further comprising:

generating a planned trajectory for an autonomous vehicle based on the predicted probability distribution.

13. The system of claim 8 , wherein the prediction error is not smaller than the threshold value, the operations further comprising:

generating a planned trajectory for an autonomous vehicle based on the predicted probability distribution; and

modifying a planned trajectory for an autonomous vehicle to account for the predicted probability distribution possibly being erroneous.

14. The system of claim 8 , wherein the anomaly estimation input is generated from the data characterizing the scene that includes the agent and the predicted probability distribution.

15. One or more non-transitory computer-readable storage media storing instructions that when executed by one or more computers cause the one or more computers to perform operations comprising:

receiving data characterizing a scene that includes an agent in an environment;

processing a behavior prediction input generated from the data using a behavior prediction neural network, wherein the behavior prediction neural network is configured to receive the behavior prediction input and to process the behavior prediction input to generate a predicted probability distribution over a plurality of possible behaviors for the agent; and

processing an anomaly estimation input generated from the data characterizing the scene that includes the agent using an anomaly estimation neural network, wherein the anomaly estimation neural network is configured to receive the anomaly estimation input and to process the anomaly estimation input to generate a prediction error for the predicted probability distribution, wherein the prediction error indicates an error between the predicted probability distribution over the plurality of possible behaviors generated by the behavior prediction neural network and another predicted probability distribution over the plurality of possible behaviors that would be generated by another behavior prediction neural network from the same behavior prediction input.

16. The non-transitory computer-readable media of claim 15 , wherein the behavior prediction neural network is an on-board behavior prediction neural network that runs on-board an autonomous vehicle, and wherein the other behavior prediction neural network is an off-board behavior prediction neural network that consumes more resources than the on-board behavior prediction neural network.

17. The non-transitory computer-readable media of claim 16 , wherein the off-board behavior prediction neural network consumes more computational resources than a compute budget allocated for behavior prediction on-board the autonomous vehicle.

18. The non-transitory computer-readable media of claim 15 , wherein the behavior prediction neural network has fewer parameters than the other behavior prediction neural network.

19. The non-transitory computer-readable media of claim 15 , wherein the prediction error is smaller than a threshold value, the operations further comprising:

generating a planned trajectory for an autonomous vehicle based on the predicted probability distribution.

20. The non-transitory computer-readable media of claim 15 , wherein the prediction error is not smaller than the threshold value, the operations further comprising:

generating a planned trajectory for an autonomous vehicle based on the predicted probability distribution; and

modifying a planned trajectory for an autonomous vehicle to account for the predicted probability distribution possibly being erroneous.

Assignments (1)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Jul 27, 2022
From: REFAAT, KHALED
To: WAYMO LLC
Reel/Frame 060641/0788 →
References Cited (16)
US 11810365B1 · Crego · 2023 [cited by examiner]
US 20190232955A1 · Grimm · 2019 [cited by examiner]
US 20190279000A1 · Pfeifle · 2019 [cited by examiner]
US 20210325901A1 · Gyllenhammar · 2021 [cited by examiner]
US 20220092456A1 · Piot · 2022 [cited by examiner]
US 20220135086A1 · Mahjourian · 2022 [cited by examiner]
US 20220171998A1 · Ni · 2022 [cited by examiner]
WO WO2018232680 · 2018 [cited by applicant]
Alahi et al., “Social 1stm: Human trajectory prediction in crowded spaces.” Proceedings of the IEEE conference on computer vision and pattern recognition. 2016, 11 pages. [cited by applicant]
Casas et al., “Spagnn: Spatially-aware graph neural networks for relational behavior forecasting from sensor data.” 2020 IEEE International Conference on Robotics and Automation (ICRA). IEEE, May 2020, 7 pages. [cited by applicant]
Hochreiter et al., “Long short-term memory.” Neural computation 9.8, Nov. 1997, 32 pages. [cited by applicant]
Huang et al., “Uncertainty-aware driver trajectory prediction at urban intersections.” Submitted on Mar. 2019, arXiv:1901.05105v2, 8 pages. [cited by applicant]
Ngiam et al., “Scene transformer: A unified architecture for predicting multiple agent trajectories.” Submitted on Mar. 2022, arXiv:2106.08417, 25 pages. [cited by applicant]
Refaat et al., “Agent prioritization for autonomous navigation.” Submitted on Sep. 2019, arXiv:1909.08792v1, 8 pages. [cited by applicant]
Varadarajan et al., “MultiPath++: Efficient Information Fusion and Trajectory Aggregation for Behavior Prediction.” submitted on Dec. 2021, arXiv: 2111.14973v3, 22 pages. [cited by applicant]
Vaswani et al., “Attention is all you need.” Advances in neural information processing systems 30, 2017, 11 pages. [cited by applicant]