MDrive: Benchmarking Closed-Loop Cooperative Driving for End-to-End Multi-agent Systems

2026-05-11

AI Summary

This work addresses the limitations of existing V2X benchmarks, which predominantly rely on open-loop evaluation and thus fail to capture the interactive dynamics and behavioral diversity inherent in real-world closed-loop cooperative driving. To bridge this gap, the authors introduce MDrive, a closed-loop cooperative driving benchmark comprising 225 high-fidelity scenarios that integrate NHTSA pre-crash typologies with real-world V2X data. Leveraging Real2Sim conversion, human-in-the-loop simulation, and a dedicated scenario generation toolbox, MDrive establishes the first end-to-end multi-agent evaluation environment that jointly ensures realism and diversity. Experimental results demonstrate that multi-agent systems consistently outperform single-agent approaches, that perception sharing does not necessarily enhance planning performance, and that negotiation mechanisms may reduce efficiency in dense, complex traffic. The complete toolchain is publicly released to support reproducible research.

Abstract

Vehicle-to-Everything (V2X) communication has emerged as a promising paradigm for autonomous driving, enabling connected agents to share complementary perception information and negotiate with each other to benefit the final planning. Existing V2X benchmarks, however, fall short in two ways: (i) open-loop evaluations fail to capture the inherently closed-loop nature of driving, leading to evaluation gaps, and (ii) current closed-loop evaluations lack behavioral and interactive diversity to reflect real-world driving. Thus, it is still unclear the extent of benefits of multi-agent systems for closed-loop driving. In this paper, we introduce MDrive, a closed-loop cooperative driving benchmark comprising 225 scenarios grounded in both NHTSA pre-crash typologies and real-world V2X datasets. Our benchmark results demonstrate that multi-agent systems are generally better than single-agent counterparts. However, current multi-agent systems still face two important challenges: (i) perception sharing enhances perceptions, but doesn't always translate to better planning; (ii) negotiation improves planning performance but harms it in complex and dense traffic scenarios. MDrive further provides an open-source toolbox for scenario generation, Real2Sim conversion, and human-in-the-loop simulation. Together, MDrive establishes a reproducible foundation for evaluating and improving the generalization and robustness of cooperative driving systems.

Problem

• V2X
• closed-loop driving
• multi-agent systems
• cooperative driving
• benchmarking

Innovation

• closed-loop cooperative driving
• multi-agent systems
• V2X communication
• benchmarking
• human-in-the-loop simulation