Autonomous Vehicle Control With Probabilistic Sensor Failure Assessment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in reliably determining the failure probability of sensor systems and ADAS subfunction commands, which affects their ability to safely transition between manual and autonomous modes.
Innovation Solution
A probabilistic method that receives unfiltered sensor signals, determines the failure probability of the sensor system and estimation signals in real time, and calculates the failure probability of ADAS subfunction commands, including the ADAS steering torque command, to autonomously control the vehicle or request driver intervention.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If the vehicle operates in autonomous mode using sensor signals, then automation level is improved, but reliability of control deteriorates due to potential sensor failures
Solution Approach 1:
The patent introduces a probabilistic monitoring system as an intermediary between the sensor system and autonomous control. This mediator calculates failure probabilities of sensor signals and ADAS commands, and uses these probabilities to determine whether to trust autonomous commands or request driver intervention, thus resolving the contradiction between automation and reliability
Solution Approach 2:
The system implements continuous feedback by monitoring sensor signal quality and ADAS command reliability in real-time. The failure probability calculations feed back into the control decision-making process, allowing the system to dynamically adjust between autonomous operation and driver takeover based on current reliability assessments
2Reliability
If the vehicle requests driver takeover frequently to ensure safety, then reliability is improved, but ease of operation deteriorates due to driver burden
Solution Approach 1:
The system uses dynamic parameter changes by adjusting the threshold for driver takeover based on contextual factors. Instead of a fixed threshold, the system considers multiple parameters including sensor failure probabilities, environmental conditions, and driving scenarios to determine when driver intervention is necessary, thereby reducing unnecessary takeovers while maintaining safety
Data Source
AI summary
A method for probabilistic autonomous vehicle control includes receiving a plurality of unfiltered sensor signals from a sensor system of a vehicle and determining, in real time, a failure probability of the sensor system using the unfiltered sensor signals. The method further includes determining a failure probability of a plurality of estimation signals at each time step using the failure probability of the sensor system and determining a failure probability of a plurality of Advanced Driver Assistance System (ADAS) subfunction commands using the failure probability of the plurality of estimation signals at each time step. Further, the method includes determining remedial actions for the plurality of ADAS subfunction commands based on the failure probability of the plurality of ADAS subfunction commands.