Autonomous Vehicle Perception Inconsistency Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current perception systems for automated vehicles lack a quality assurance mechanism at runtime, leading to potential catastrophic errors in shared spaces with humans, as they are imperfect and based on non-explainable architectures, making it challenging to assess and ensure the reliability of perceived information without ground-truth data.
Innovation Solution
A computer-implemented method that analyzes the evolution of the driving scene by comparing new perception results with past ones, calculating boundaries of possible future states based on object behavior assumptions, and sending notifications to the electronic control unit when inconsistencies are detected, allowing for potential safety actions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If perception systems use non-explainable architectures like deep neural networks, then detection accuracy is improved, but quality assurance and reliability assessment become difficult
Solution Approach 1:
The patent introduces an intermediary consistency check module that acts as a mediator between the perception system and the control system. This module verifies the consistency of perception outputs without needing to understand the internal workings of the neural network, thus maintaining detection accuracy while enabling quality assurance through behavioral consistency verification.
Solution Approach 2:
The system implements feedback by continuously monitoring perception outputs and comparing them against expected behavioral patterns. When inconsistencies are detected, the system provides feedback to the control unit to adjust or reject decisions, thereby ensuring reliability assessment without compromising the neural network's detection accuracy.
2Device complexity
If no ground-truth information is available at runtime, then system complexity is reduced, but quality assessment of perception systems becomes impossible
Solution Approach 1:
The system performs self-assessment by generating its own reference framework through propagation of past results. Instead of relying on external ground-truth data, the system uses its historical performance and behavioral assumptions to create internal consistency checks, enabling quality assessment without increasing system complexity or requiring external supervision.
Solution Approach 2:
The system performs preliminary propagation of past perception results to establish expected future states before actual new observations arrive. This preliminary action creates a reference framework that enables subsequent consistency verification, allowing quality assessment without needing ground-truth information or increasing system complexity.
3Productivity
If perception errors are not detected, then system operation continues smoothly, but catastrophic errors may occur in shared spaces
Solution Approach 1:
The system applies preliminary anti-action by proactively detecting potential perception errors through consistency checks before they can lead to catastrophic failures. The propagation mechanism predicts expected states and flags deviations in advance, preventing harmful actions while maintaining smooth operation through early error detection and control unit intervention.
4Speed
If new perception results are immediately processed, then response time is improved, but consistency with past observations cannot be verified
Solution Approach 1:
The system performs preliminary propagation of past results to establish expected states before processing new observations. This preliminary action enables rapid consistency verification without delaying the processing of new perception results, maintaining fast response time while ensuring reliability through pre-computed reference frameworks.
Data Source
AI summary
A system and method for the detection of inconsistencies in perception systems of autonomous vehicles is described. The system receives the observations of objects in the surrounding environment from one or more sensors or perception systems of an automated vehicle. At actual time, the system estimates the consistency of the currently observed elements of the perception system according to the previous inputs received. This consistency is decided by calculating the boundaries of possible states of the previously observed elements, based on the received information and on assumptions.