Autonomous Vehicle Sensor Fault Detection via Secondary Occupancy Model
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Advanced driver assistance systems (ADAS) and autonomous driving systems face faults due to faulty sensors or aberrant sensor data processing, leading to incorrect object detection and potentially hazardous automated safety and driving functionalities.
Innovation Solution
A self-diagnosis system that generates an environmental model from sensor data and compares it with a secondary occupancy space data structure to identify sensor faults, estimating the time before a potential crash and recalibrating sensors as needed to prevent errors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a primary computing system generates an environmental model from sensor data for autonomous driving, then object detection capability is improved, but system reliability deteriorates due to potential sensor faults and aberrant data processing
Solution Approach 1:
A secondary computing system acts as an intermediary to generate an occupancy space data structure from sensor data, which then serves as a reference for fault detection. This intermediary structure enables comparison between the primary environmental model and an independently generated representation, allowing fault identification without directly modifying the primary detection pipeline.
Solution Approach 2:
The system implements feedback by comparing the environmental model with the occupancy space data structure and using the discrepancies to identify sensor faults. The fault identification results can trigger recalibration actions, creating a closed-loop feedback mechanism that continuously monitors and corrects sensor performance to maintain system reliability.
2Reliability
If the system implements fault detection by comparing environmental model with occupancy space data structure, then sensor fault identification is improved, but computational complexity increases
Solution Approach 1:
The computational task is segmented into two independent processes: generating the environmental model and generating the occupancy space data structure. This segmentation allows parallel processing and independent optimization of each data structure generation algorithm, reducing the overall computational burden compared to a single complex verification system.
Solution Approach 2:
Instead of creating a complex verification system, the patent creates a simplified copy or representation of the sensor data processing pipeline that generates the occupancy space data structure. This copy uses the same sensor inputs but produces a different data representation that can be efficiently compared for fault detection.
3Reliability
If the system estimates time before crash based on vehicle velocity and free space, then crash prevention capability is improved, but response time is reduced due to real-time calculation requirements
Solution Approach 1:
The system performs preliminary action by continuously estimating the time to crash based on current vehicle velocity and free space measurements. This ongoing estimation allows the system to identify potentially hazardous situations before they become critical, providing advance warning and enabling preventive actions to be taken with sufficient time margin.
Data Source
AI summary
This application discloses self-diagnosis of faults for an assisted or automated driving system of a vehicle. A primary computing system can generate an environmental model populated with objects detected from measurement data from sensors. A secondary computing system can generate a secondary system data structure configured to identify objects located around the vehicle that were detected from measurement data from the sensors. The secondary computing system can identify a fault in the sensors based on a comparison of the secondary system data structure with the environmental model. The secondary computing system also can estimate an amount of time before the vehicle crashes from a vehicle velocity vector and free space in the environment around the vehicle, and identify the fault in the sensors based on the estimated amount of time before the vehicle crashes.