Overview
An autonomous vehicle was deployed for operation on public roads within a defined Operational Design Domain (ODD), using a multi-sensor perception stack including cameras, radar, and lidar. The system was designed to operate driverless in complex and dynamic environments. 

A key focus of this case was managing sensor degradation caused by environmental conditions such as dust, mud, rain, or water accumulation, which can reduce perception performance while operating within the intended ODD. 

Challenges
The primary challenge was identifying and managing environmental conditions that cause degradation of onboard sensors while the vehicle is operating within its defined ODD. This required systematic identification of conditions that transition the system into degraded perception states that could result in scenarios outside of the defined ODD. 

Given the dynamic nature of public road operation, the system needed to detect changes in sensor performance in real time and ensure appropriate operational responses to maintain safe system behavior. All identified degradation conditions were formally characterized as ODD boundary-related constraints. 

Considerations included balancing operational autonomy, safety assurance, and human oversight requirements: 

• Additional operational resources were required to monitor sensor health and perception capability in real time via remote operations 

• Increased cognitive load and responsibility on remote operators, requiring task definition, documentation, and training 

• Maintaining driverless operation while introducing fallback and monitoring mechanisms that could interrupt or modify autonomy behavior 

These considerations directly influenced system design for remote supervision and intervention strategies. 

Approach
A structured safety and operational mitigation strategy was implemented: 

• Identification and classification of environmental conditions leading to sensor degradation (e.g., occlusion from dust, mud, rain, or water) 

• Formalization of these conditions as ODD boundary violations or near-boundary states 

• Development of real-time detection mechanisms for sensor degradation and environmental changes 

• Implementation of operational mitigations to adjust AV behavior dynamically under degraded perception conditions 

• Integration of remote operations procedures to supervise and support system behavior when degradation is detected 

• Validation of mitigations through controlled closed-course testing prior to deployment 

Key Insights
The project highlighted several important findings:

• Sensor degradation must be explicitly treated as an ODD boundary condition, not only as a hardware reliability issue 

• Real-time monitoring of perception quality is essential for safe autonomous operation in unstructured environments 

• Remote operations introduce significant human-factor considerations, including cognitive load and task complexity that must be formally managed 

• Mitigation strategies must be fully validated, not only defined, before deployment into safety-critical autonomous systems 

• Environmental robustness is a key determinant of safe driverless operation, especially in public road scenarios 

Outcome
Sensor degradation risks were formally documented and characterized within the system’s safety concept. Mitigation strategies were developed, implemented, and translated into operational procedures and remote operator training material. These mitigations were validated to ensure safe system response under realistic degradation scenarios. 

Following validation, the driverless autonomous vehicle system was successfully deployed for operation on public roads with integrated monitoring and mitigation strategies in place.