REACT: Runtime-Enabled active collision-avoidance technique for autonomous driving

Huang, Heye; Cheng, Hao; zhou, zhiyuan; Wang, Zijin; Wang, Haoran; Liu, Qi; Li, Xiaopeng · 2025 · Advanced Engineering Informatics

DOI: 10.1016/j.aei.2025.104248

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper addresses the challenge of achieving rapid, effective active collision avoidance in dynamic, mixed-traffic environments for autonomous driving. Existing risk assessment systems often rely on rule-based heuristics or offline modeling, which struggle to capture the unpredictability of multi-agent interactions and lack real-time, closed-loop validation. To overcome these limitations, the authors propose REACT (Runtime-Enabled Active Collision-avoidance Technique), a lightweight, closed-loop framework that integrates runtime risk assessment with active avoidance control. The system aims to proactively identify high-risk zones and generate feasible, interpretable avoidance behaviors by leveraging energy transfer principles and human-vehicle-road interaction modeling. The REACT framework operates through a three-stage logic: risk modeling, grid-based risk map generation, and active avoidance. It constructs a multi-source risk field based on kinetic energy, velocity differences, and mass, modeling interactions between the ego vehicle and surrounding agents as potential fields analogous to electrostatic forces. This field incorporates directional sensitivity using elliptical distance metrics and accounts for road boundary constraints via spring potential energy models. The continuous risk field is discretized into a two-dimensional grid centered on the ego vehicle to compute a unified runtime risk value and identify dominant risk directions. Based on normalized risk values and adaptive thresholds, the system triggers a hierarchical warning strategy (Level 0: Safe, Level 1: Warning, Level 2: Emergency) to execute appropriate control actions, such as deceleration or emergency braking. Evaluations were conducted across four representative high-risk scenarios: car-following braking, cut-in, rear-approaching, and intersection conflicts. The results demonstrate that REACT achieves 100% safe avoidance with zero false alarms or missed detections. The system exhibits strong real-time performance with latency under 50 ms and aligns closely with human driver cognition, providing warning lead times of less than 0.4 seconds. Compared to traditional metrics like Time to Collision (TTC) and Responsibility-Sensitive Safety (RSS), which are limited by single-dimensional assessments or scenario-specific tuning, REACT offers superior generalization and accuracy in complex, dynamic environments. The significance of this work lies in its ability to provide a computationally lightweight, runtime-compatible solution for safety-critical autonomous systems. By unifying behavior intention and physical modeling into a directional risk field, REACT enables accurate risk prediction and robust avoidance control without the high computational complexity typical of field-based methods. The study concludes that REACT’s state-of-the-art accuracy and low-latency execution highlight its potential for real-time deployment in embedded in-vehicle platforms, addressing the critical need for proactive safety architectures in autonomous driving.

Key finding

The REACT framework achieves 100% safe avoidance with zero false alarms or missed detections across four high-risk scenarios while maintaining a latency of less than 50 ms.

Methodology

simulation_modeling

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed. Discovered via author_sweep_intake on 2026-05-28.

StageOutcomeToolModelPromptAttemptsCompleted
discover success author_sweep 2 2026-05-28
archive success canonical_url 1 2026-06-06
extract success cached 3 2026-06-10
clean success clean 1 2026-06-07
chunk success chunk 1 2026-06-07
embed success embed Qwen/Qwen3-Embedding-8B 1 2026-06-07
enrich failed 5 2026-07-02
promote success 1 2026-06-04
summarize success llm qwen3.6-27b-prismaquant summ-v5 2 2026-06-10
tag success vector_similarity 15 2026-06-11
verify success 2 2026-06-10

Summary generated by qwen3.6-27b-prismaquant on 2026-06-10; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).