Autonomous Vehicle Safety Model for Noisy Sensor Conditions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current autonomous vehicle safety models, such as the 'Responsible Sensitive Safety' (RSS) model, assume a perfect sensing environment and may fail to guarantee safe behavior in scenarios with noisy sensor data, leading to potential accidents, especially in bidirectional single-lane roads or intersections where vehicle positioning is uncertain.
Innovation Solution
An improved safety model and associated rules are introduced to minimize the impact of noisy sensor readings by decoupling sensing mechanisms from planning phases and incorporating virtual objects to define road boundaries, ensuring safe vehicle operation even with imperfect sensing data, and limiting acceleration to prevent dangerous collisions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the autonomous vehicle uses a baseline safety model assuming perfect sensing, then the safety model is simple and easy to implement, but it fails to guarantee safe behavior in noisy sensor environments
Solution Approach 1:
The safety model is segmented into distinct modules: a baseline safety model for normal operation and an extended safety model for noisy sensor conditions. Each module handles specific scenarios, allowing the system to maintain simplicity when possible while providing enhanced safety when needed.
Solution Approach 2:
The system performs preliminary sensing to detect potentially noisy conditions before making safety decisions. By identifying noisy sensor environments in advance, the system can switch to the extended safety model proactively, ensuring safety guarantees are maintained without unnecessarily complicating normal operation.
2Adaptability or versatility
If the autonomous vehicle operates in bidirectional single-lane roads or intersections, then the vehicle can access more road scenarios, but the uncertainty in vehicle positioning increases due to noisy sensor data
Solution Approach 1:
The extended safety model proactively compensates for positioning uncertainty by establishing larger safety margins and adjusted distance thresholds before conflicts occur. This preliminary anti-action counteracts the effects of noisy positioning data, allowing the vehicle to safely operate in complex scenarios like bidirectional roads and intersections despite measurement imprecision.
3Reliability
If the autonomous vehicle limits acceleration to prevent dangerous collisions, then collision risk is reduced, but the vehicle's ability to respond to urgent situations is constrained
Solution Approach 1:
The acceleration limiting is implemented dynamically rather than as a fixed constraint. The system adjusts acceleration limits in real-time based on the noisy condition flags and environmental context, allowing higher acceleration when safety margins permit and stricter limits when uncertainty is high, thus balancing collision prevention with responsive capability.
Data Source
AI summary
Various systems and methods for controlling a vehicle using driving policies are described herein. A system for controlling a vehicle using driving policies includes a memory device; and a processor subsystem to access instructions on the memory device that cause the processor subsystem to: operate a host vehicle using a driving policy from a policy repository, the host vehicle operating in a lane on a first road, the driving policy governed by a safety model; detect a second vehicle, the second vehicle operating in a second lane; determine whether the second vehicle is an oncoming vehicle or an intersecting vehicle, the oncoming vehicle operating on the first road with the first and second lanes in adjacent bidirectional arrangement, and the intersecting vehicle operating on a second road that intersects the first road; and initiate a vehicle maneuver of the host vehicle to reduce or avoid a collision with the second vehicle, based on the safety model, the vehicle maneuver performed based on whether the second vehicle is an oncoming vehicle or an intersecting vehicle.