Defer to Plan: Adaptive Multi-Agent Fusion for End-to-End V2X Driving

2026-07-22

AI Summary

This work addresses the limitation of existing V2X cooperative perception methods, which optimize perception performance in isolation and fail to effectively support downstream planning and control tasks. To bridge this gap, we propose the first end-to-end cooperative driving system that jointly optimizes perception and planning while directly aligning with driving decision objectives. Our approach introduces a MotionNetwork to model temporal dynamics, integrates an attention mechanism to compress spatial features, and employs an autoregressive decoder for adaptive multi-agent feature fusion. Notably, we pioneer the application of a Mixture-of-Experts (MoE) architecture in V2X settings to enhance representation capacity for heterogeneous features. In closed-loop evaluation, the system achieves a driving score of 79.72, surpassing the current state-of-the-art CoDriving baseline by 3.33% while maintaining communication efficiency.

Abstract

Vehicle-to-everything-aided autonomous driving (V2X-AD) significantly enhances driving performance through information sharing. However, existing collaborative perception methods only optimize module-level perception capabilities and fail to effectively serve the ultimate planning and control tasks. We propose an end-to-end collaborative driving system that directly optimizes planning task performance. The system employs MotionNetwork to fuse historical temporal information, utilizes attention mechanisms to efficiently compress spatial features into compact tokens, and adaptively fuses multi-agent features through an autoregressive decoder. Additionally, we introduce Mixture-of-Experts (MoE) architecture to enhance the model's representation capacity for heterogeneous features. Experiments demonstrate that our method achieves a driving score of 79.72, surpassing the state-of-the-art CoDriving baseline (77.15) by 3.33% in closed-loop evaluation while maintaining communication efficiency.

Problem

• V2X-AD
• collaborative perception
• planning task
• multi-agent fusion
• end-to-end driving

Innovation

• end-to-end V2X driving
• adaptive multi-agent fusion
• Mixture-of-Experts
• autoregressive decoder
• attention-based token compression