WorkDrive: Roadwork Chain of Causation for Autonomous Driving

2026-07-16

AI Summary

This work addresses the challenge of accurate motion planning for autonomous driving in road construction zones, where conventional visual cues are often absent. To tackle this issue, the authors propose a Construction-zone Chain-of-Causality (CoC) annotation framework that establishes a perception-driven, structured causal reasoning pathway, directing the model’s attention to critical elements such as temporary traffic devices and aligning its reasoning with trajectory prediction. The approach integrates a multi-task perception pipeline, supervised fine-tuning, and a GRPO reinforcement learning strategy grounded in lateral meta-action consistency to jointly optimize reasoning and planning. Evaluated on the ROADWork dataset, the method reduces the average displacement error (ADE) of predicted trajectories by 12.0% over the baseline, with CoC contributing a 9.0% improvement and GRPO providing an additional 3.0% gain.

Abstract

Autonomous driving vision-language models (VLMs) struggle in roadwork zones, where familiar visual cues such as lane markings and permanent signs are altered or absent, and temporary devices such as cones and barriers redefine the drivable corridor. VLMs can detect these objects, but without explicit guidance they anchor their reasoning on familiar elements from pre-training and fail to connect work-zone observations to correct planning decisions. We propose WorkDrive, a framework that constructs perception-grounded causal reasoning for work zones and aligns it with trajectory prediction. An automated multitask perception pipeline extracts structured scene facts and injects them into a Chain-of-Causation (CoC) annotation pipeline, redirecting the annotator's attention to domain-specific elements. The resulting reasoning labels are used for supervised fine-tuning, followed by reinforcement learning with a single reward: consistency between lateral meta-actions and the predicted trajectory. On ROADWork, the largest public work-zone dataset, the proposed roadwork CoC reduces trajectory average displacement error (ADE) by 9.0\%, and consistency-based GRPO yields a further 3.0\%, achieving progressive improvement over the trajectory-only baseline. Code and data will be publicly released.

Problem

• autonomous driving
• roadwork zones
• vision-language models
• causal reasoning
• trajectory prediction

Innovation

• Chain-of-Causation
• perception-grounded reasoning
• work-zone driving
• trajectory consistency
• vision-language models