Perception Error Modeling for Faster Autonomous Vehicle Simulation

Solution Overview

Problem

The computational and time-intensive nature of simulating multiple scenarios for testing autonomous systems, such as autonomous vehicles, leads to significant resource usage and inefficiencies in determining performance metrics, particularly when variations in scenarios need to be exhaustively tested.

Innovation Solution

The use of a perception error model to expedite the determination of performance metrics by modifying simulated object detections and reducing the number of required simulations, allowing for quicker estimation of overall vehicle performance and component tuning.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If multiple iterations of multiple scenarios are run to thoroughly test autonomous systems, then measurement precision and reliability are improved, but computational use, time, memory, and energy increase by multiple orders of magnitude

Engineering Contradiction:

Improveperformance metric accuracyVSAvoidsimulation time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent pre-computes and stores error distributions from perception systems for various object classes and scenarios during an offline training phase. These pre-computed error characteristics are then reused during online simulation without requiring repeated full perception pipeline executions, significantly reducing simulation time while maintaining accuracy.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent creates simplified copies of the perception system's error characteristics through statistical models and distributions. Instead of running the full perception system multiple times, it uses copied error data from training scenarios to simulate perception outcomes in test scenarios, reducing computational load while preserving measurement precision.

Inventive Principle:

Principle #26Copying

2Measurement precision

If multiple iterations of multiple scenarios are run to thoroughly test autonomous systems, then measurement precision and reliability are improved, but computational use, time, memory, and energy increase by multiple orders of magnitude

Engineering Contradiction:

Improveperformance metric accuracyVSAvoidsimulation throughput

Core Design Contradiction:

Measurement precisionVSProductivity

Solution Approach 1:

The patent pre-computes and stores error distributions from perception systems for various object classes and scenarios during an offline training phase. These pre-computed error characteristics are then reused during online simulation without requiring repeated full perception pipeline executions, significantly reducing simulation time while maintaining accuracy.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent creates simplified copies of the perception system's error characteristics through statistical models and distributions. Instead of running the full perception system multiple times, it uses copied error data from training scenarios to simulate perception outcomes in test scenarios, reducing computational load while preserving measurement precision.

Inventive Principle:

Principle #26Copying

3Reliability

If multiple iterations of multiple scenarios are run to thoroughly test autonomous systems, then reliability is improved, but computational use and energy increase by multiple orders of magnitude

Engineering Contradiction:

Improvesystem safety validationVSAvoidcomputational energy

Core Design Contradiction:

ReliabilityVSUse of energy by moving object

Solution Approach 1:

The patent pre-computes and stores error distributions from perception systems for various object classes and scenarios during an offline training phase. These pre-computed error characteristics are then reused during online simulation without requiring repeated full perception pipeline executions, significantly reducing simulation time while maintaining accuracy.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent creates simplified copies of the perception system's error characteristics through statistical models and distributions. Instead of running the full perception system multiple times, it uses copied error data from training scenarios to simulate perception outcomes in test scenarios, reducing computational load while preserving measurement precision.

Inventive Principle:

Principle #26Copying

Data Source

PatentUS12055941B1Perception error model for fast simulation and estimation of perception system reliability and/or for control system tuning

Publication Date: 2024.08.06 ZOOX INC

AI summary

An error estimation component may generate contours or a heat map of a likelihood that an object detection or prediction generated by a perception component of an autonomous vehicle is inaccurate. The contours or heat map may be used to lock or unlock actions available to a planning component and the planning component may use the contours or heat for determining a trajectory for controlling the autonomous vehicle. The planning component may additionally or alternatively alter operation of a sensor or determine a trajectory to unlock a target action that is currently locked due to a size/extent of uncertainty indicated by a contour/heat map.