VLALight: A Vision-Language-Action Model for Traffic Signal Control

2026-09-29

AI Summary

This work addresses the disconnection between visual observation and decision-making in existing traffic signal control methods, which typically rely on handcrafted states or isolated perception modules. We propose the first end-to-end Vision-Language-Action (VLA) model for traffic signal control that directly maps multi-view video inputs to coordinated signal actions. The model introduces a novel VLA architecture incorporating an adaptive fast-slow reasoning mechanism to balance decision quality with computational cost. Training integrates two-stage supervised cold-start initialization, cooperative multi-agent reinforcement learning, and topology-aware collaboration techniques. Extensive experiments across seven real-world datasets demonstrate that our approach significantly outperforms existing baselines, validating the potential of the VLA paradigm for physical traffic control applications.

Abstract

Traffic signal control (TSC) is essential for improving urban mobility and reducing congestion. Although roadside cameras are widely deployed at signalized intersections and provide rich visual observations of evolving traffic, existing TSC methods typically rely on manually engineered traffic states or separate perception modules, creating a gap between physical observations and control decisions. We present VLALight, the first vision-language-action (VLA) model for end-to-end traffic signal control from multi-view roadside videos. VLALight directly maps visual observations to coordinated signal actions through multi-target spatiotemporal traffic reasoning and topology-aware cooperative perception across intersections. To establish this capability, we develop a two-stage supervised cold-start training strategy for visual traffic understanding and signal decision-making, followed by cooperative agentic reinforcement learning that jointly optimizes local control and network-wide traffic efficiency. Furthermore, VLALight introduces adaptive fast and slow reasoning modes, enabling the policy to allocate deeper reasoning only when additional deliberation provides sufficient control benefits. Through balanced mode-aware rollouts and relative advantage optimization, VLALight learns to trade off decision quality and inference cost. Extensive experiments on seven real-world traffic-flow datasets across three urban networks demonstrate that VLALight consistently outperforms transportation-based, RL-based, and LLM/VLM-based baselines. Ablation studies validate the effectiveness of cooperative perception, network-level optimization, and adaptive reasoning. These results demonstrate the potential of VLA models for real-world physical traffic control. Our project is available at https://github.com/usail-hkust/VLALight.git.

Problem

• Traffic Signal Control
• Vision-Language-Action Model
• End-to-End Control
• Cooperative Perception

Innovation

• Vision-Language-Action Model
• Traffic Signal Control
• Cooperative Perception
• Agentic Reinforcement Learning
• Adaptive Reasoning