Dimensions overall score 4.0

GitHub Code Pulse

No public code linked for this paper yet.

• The Lateral Evasion Acceleration (LEA) metric quantifies the minimum steering effort required to avoid a predicted collision, adapted from prior lateral evasion kinematics.method 90%
• A reachability-based ellipsoidal collision filter is used to ensure only dynamically plausible threats are scored by the metrics.technical 90%
• Evaluation on nuScenes and Argoverse 2 datasets shows that 65-93% of perception errors are non-critical.result 85%
• Spearman correlation analysis confirms that the three proposed effort-based metrics capture safety-relevant information inaccessible to established time-based, deceleration-based, or normalized criticality measures.result 85%
• The proposed framework enables targeted mining of the most critical perception failures.method 80%
• The conservatism of the ellipsoidal reachability filter can flag objects not in the ego's lane as potential threats.limitation 80%
• The proposed False Speed Reduction (FSR) metric quantifies the cumulative unwarranted velocity loss caused by persistent false-positive perception errors.method 95%
• The proposed Maximum Deceleration Rate (MDR) metric quantifies the peak braking demand caused by false-negative perception errors under a constant-acceleration model.method 95%

Competitive landscape

Competitor map is still being generated for this paper. Enable generation or check back soon.

Startup potential card