Method, system, and media for calculating a surprise score by comparing a posterior state to a prior probability distribution

Patent metadata

Field	Value
Publication number	US-12576889-B2
Application number	US-202418672812-A
Country	US
Kind code	B2
Filing date	May 23, 2024
Priority date	Aug 11, 2021
Publication date	Mar 17, 2026
Grant date	Mar 17, 2026

Abstract

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for computing a backward looking surprise metric for autonomously driven vehicles. One of the methods includes obtaining first data representing one or more previously predicted states of an agent along one or more predicted trajectories of the agent at a first time step. Second data representing one or more states of the agent at a subsequent time step is obtained. A surprise score is computed from a measure of a difference between the first data computed for the one or more predicted trajectories for the prior time step and the second data computed for the one or more predicted states for the subsequent time step.

First claim

The invention claimed is:

1 . A method comprising:
obtaining first data comprising a prior probability distribution representing a plurality of previously predicted states of an agent along one or more predicted trajectories of the agent from a first time step, wherein the plurality of previously predicted states were generated based on a prior state of the agent at a prior time step that occurred before a current time step; obtaining second data comprising a posterior state representing one or more states of the agent at a subsequent time step; and computing a surprise score for the agent by comparing the posterior state to the prior probability distribution, wherein computing the surprise score for the agent comprises computing a likelihood of the posterior state based on the prior probability distribution.

2 . The method of claim 1 , wherein computing the surprise score for the agent comprises computing a measure of a difference between a predicted state of the agent and an actual state of the agent, wherein the first data comprises the predicted state and wherein the second data comprises the actual state.

3 . The method of claim 1 , wherein computing the surprise score for the agent comprises computing a measure of a difference between a posterior probability distribution and the prior probability distribution over states, wherein the second data comprises the posterior probability distribution over states.

4 . The method of claim 1 , wherein computing the surprise score for the agent comprises computing a measure of a difference between a first mixture model representing the plurality of previously predicted states and a second mixture model representing the one or more states of the agent at the subsequent time step, wherein the first data comprises the first mixture model, the second data comprises the second mixture model.

5 . The method of claim 1 , wherein obtaining the second data comprises computing a predicted state of the agent at the subsequent time step along a plurality of predicted trajectories based on a current state of the agent.

6 . The method of claim 1 , further comprising performing a parameter sweeping process to generate multiple surprise scores using only a single inference pass of a generative model that generates predicted states of the agent, wherein a first parameter of the parameter sweeping process is which time step is used as the prior time step, and wherein a second parameter of the parameter sweeping process is a number of steps to predict into a future time step.

7 . The method of claim 1 , wherein the method is performed by a computer-system on-board an autonomous or a semi-autonomous vehicle.

8 . The method of claim 1 , wherein the method is performed by a computer-system in an offline status in a location different from an autonomous or a semi-autonomous vehicle.

9 . A system comprising:
one or more computers and one or more storage devices storing instructions that are operable, when executed by the one or more computers, to cause the one or more computers to perform operations comprising:
obtaining first data comprising a prior probability distribution representing a plurality of previously predicted states of an agent along one or more predicted trajectories of the agent from a first time step, wherein the plurality of previously predicted states were generated based on a prior state of the agent at a prior time step that occurred before a current time step;
obtaining second data comprising a posterior state representing one or more states of the agent at a subsequent time step; and
computing a surprise score for the agent by comparing the posterior state to the prior probability distribution, wherein computing the surprise score for the agent comprises computing a likelihood of the posterior state based on the prior probability distribution.

10 . The system of claim 9 , wherein computing the surprise score for the agent comprises computing a measure of a difference between a predicted state of the agent and an actual state of the agent, wherein the first data comprises the predicted state and wherein the second data comprises the actual state.

11 . The system of claim 9 , wherein computing the surprise score for the agent comprises computing a measure of a difference between a posterior probability distribution and the prior probability distribution over states, wherein the second data comprises the posterior probability distribution over states.

12 . The system of claim 9 , wherein computing the surprise score for the agent comprises computing a measure of a difference between a first mixture model representing the plurality of previously predicted states and a second mixture model representing the one or more states of the agent at the subsequent time step, wherein the first data comprises the first mixture model, the second data comprises the second mixture model.

13 . One or more non-transitory computer storage media encoded with computer program instructions that when executed by one or more computers cause the one or more computers to perform operations comprising:
obtaining first data comprising a prior probability distribution representing a plurality of previously predicted states of an agent along one or more predicted trajectories of the agent from a first time step, wherein the plurality of previously predicted states were generated based on a prior state of the agent at a prior time step that occurred before a current time step; obtaining second data comprising a posterior state representing one or more states of the agent at a subsequent time step; and computing a surprise score for the agent by comparing the posterior state to the prior probability distribution, wherein computing the surprise score for the agent comprises computing a likelihood of the posterior state based on the prior probability distribution.

14 . The non-transitory computer storage media of claim 13 , wherein computing the surprise score for the agent comprises computing a measure of a difference between a predicted state of the agent and an actual state of the agent, wherein the first data comprises the predicted state and wherein the second data comprises the actual state.

15 . The non-transitory computer storage media of claim 13 , wherein computing the surprise score for the agent comprises computing a measure of a difference between a posterior probability distribution and the prior probability distribution over states, wherein the second data comprises the posterior probability distribution over states.

16 . The non-transitory computer storage media of claim 13 , wherein computing the surprise score for the agent comprises computing a measure of a difference between a first mixture model representing the plurality of previously predicted states and a second mixture model representing the one or more states of the agent at the subsequent time step, wherein the first data comprises the first mixture model, the second data comprises the second mixture model.

17 . The non-transitory computer storage media of claim 13 , wherein obtaining the second data comprises computing a predicted state of the agent at the subsequent time step along a plurality of predicted trajectories based on a current state of the agent.

18 . The non-transitory computer storage media of claim 13 , wherein the operations further comprise:
performing a parameter sweeping process to generate multiple surprise scores using only a single inference pass of a generative model that generates predicted states of the agent, wherei

Assignees

• Waymo Llc

Inventors

• Dinparastdjadid Azadeh
• Engstrom Johan
• Chen Haoyu
• Supeene Isaac
• Wang Menghui

Classifications

• B60W2554/404

  Characteristics · CPC title

• B60W2554/4026

  Cycles · CPC title

• B60W2554/4029

  Pedestrians · CPC title

• G07C5/0808 Primary

  Diagnosing performance data (testing of vehicles G01M17/00; testing of electrical installation on vehicles G01R31/005) · CPC title

• B60W60/0027 Primary

  using trajectory prediction for other traffic participants · CPC title