Minimize Sensor Camera Degradation on Advanced Driver Assistance Systems to Maintain Optimal Lane Line Detection.
Study Assessed the Impact of Degradations on Sensor Performance Through In-Vehicle Testing.
Virginia
Safety Implications of Potential Advanced Driver Assistance Systems Sensor Degradation.
Summary Information
Advanced driver assistance systems (ADAS) help human drivers operate vehicles with technologies such as forward collision warning, automatic emergency braking, adaptive cruise control, blind spot warning, lane departure warning, lane keep assist, and surround view cameras. This project’s goal was to increase understanding of the operational, performance, reliability, and maintenance issues associated with ADAS sensors, and their gradual degradation over time. The researchers developed simulated degradations (e.g., scratches, dirt and debris buildup, occlusions such as sap or tar, etc.) and evaluated different sensor types and models.
Using vehicles with degraded and non-degraded ADAS sensors (cameras and radar), the study conducted a series of crash avoidance and driver assistance tests. Sensor degradations were evaluated at three severity levels (low, medium, and high). To support data collection in dynamic, target-vehicle scenarios, the test vehicle was instrumented with camera, radar, and LiDAR sensors, along with a ground-truth system. Sensor data collected under degraded conditions were then systematically compared with baseline, non-degraded measurements. The following were some lessons learned from that process.
- Minimize sensor camera degradation to maintain optimal lane line detection. The testing found that lane line detection was particularly impacted by green tint, shellac, and sandblasted degradations.
- Recognize the potential risk of inflated performance from non-embedded perception algorithms. Reliance on external ground truth data, such as the Differential Global Positioning System (DGPS), likely improved object detection beyond what onboard-only perception systems can achieve. This discrepancy limits the comparability of performance metrics across systems.
- Consider the buffering effect of sensor fusion in maintaining ADAS functionality. The system-level tests showed that Radar degradations did not prevent Adaptive Cruise Control (ACC) from engaging on any of the test vehicles. This suggests that camera-radar fusion and the longer detection timelines in ACC scenarios help compensate for sensor degradation, but isolated sensor testing is needed to assess individual input vulnerabilities.
- Prioritize signal strength metrics alongside positional accuracy. This study found sensor degradations affected radar cross-section (RCS) intensity more than positional accuracy. This suggests that while object position detection may remain stable, signal degradation could lead to delayed or missed detections.
