Effort-Based Criticality Metrics for Evaluating 3D Perception Errors in Autonomous Driving

2026-03-30

AI Summary

Existing criticality metrics, such as Time-to-Collision (TTC), struggle to differentiate the distinct safety impacts of false positives and false negatives in 3D perception for autonomous driving. This work proposes three novel driving-effort-based criticality measures: False-positive-induced Speed Reduction (FSR), Missed-detection-induced Maximum Deceleration Requirement (MDR), and Lateral Evasion Acceleration (LEA)—the first to directly link perception errors to vehicle control effort. By integrating constant-acceleration motion models, reachability ellipsoid-based collision filtering, and frame-level matching with trajectory-level aggregation, the method effectively filters out dynamically infeasible threats. Experiments on nuScenes and Argoverse 2 demonstrate that 65–93% of perception errors are non-critical, and the proposed metrics exhibit strong correlation with safety; Spearman analysis confirms they capture critical risk information overlooked by conventional metrics.

Abstract

Criticality metrics such as time-to-collision (TTC) quantify collision urgency but conflate the consequences of false-positive (FP) and false-negative (FN) perception errors. We propose two novel effort-based metrics: False Speed Reduction (FSR), the cumulative velocity loss from persistent phantom detections, and Maximum Deceleration Rate (MDR), the peak braking demand from missed objects under a constant-acceleration model. These longitudinal metrics are complemented by Lateral Evasion Acceleration (LEA), adapted from prior lateral evasion kinematics and coupled with reachability-based collision timing to quantify the minimum steering effort to avoid a predicted collision. A reachability-based ellipsoidal collision filter ensures only dynamically plausible threats are scored, with frame-level matching and track-level aggregation. Evaluation of different perception pipelines on nuScenes and Argoverse~2 shows that 65-93% of errors are non-critical, and Spearman correlation analysis confirms that all three metrics capture safety-relevant information inaccessible to established time-based, deceleration-based, or normalized criticality measures, enabling targeted mining of the most critical perception failures.

Problem

• criticality metrics
• 3D perception errors
• autonomous driving
• false-positive
• false-negative

Innovation

• effort-based criticality
• False Speed Reduction
• Maximum Deceleration Rate
• Lateral Evasion Acceleration
• reachability-based filtering