Revisit Mixture Models for Multi-Agent Simulation: Experimental Study within a Unified Framework

2025-01-28

AI Summary

To address insufficient behavioral realism, high prediction errors, and severe distribution shift in multi-agent modeling for autonomous driving simulation, this paper proposes UniMM—a unified mixed-modality modeling framework. UniMM supports joint continuous-discrete modeling and introduces three novel variants: discrete-only, anchor-free, and anchor-based. To mitigate train-deploy distribution shift, UniMM incorporates a closed-loop sampling mechanism, synergistically combined with forward-component matching and multi-step predictive modeling to suppress shortcut learning and off-policy bias. Evaluated on the WOSAC benchmark, UniMM achieves state-of-the-art (SOTA) performance across all variants, consistently outperforming existing methods. Empirical results demonstrate that closed-loop sampling is pivotal for enhancing behavioral fidelity in simulation.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Multiagent Systems: Adversarial Agents
• Search and Optimization: Sampling/Simulation-based Search

Application Category

• User Modeling, Personalization and Recommendation: User modeling and simulation for interactive and conversational systems
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

Simulation plays a crucial role in assessing autonomous driving systems, where the generation of realistic multi-agent behaviors is a key aspect. In multi-agent simulation, the primary challenges include behavioral multimodality and closed-loop distributional shifts. In this study, we revisit mixture models for generating multimodal agent behaviors, which can cover the mainstream methods including continuous mixture models and GPT-like discrete models. Furthermore, we introduce a closed-loop sample generation approach tailored for mixture models to mitigate distributional shifts. Within the unified mixture model~(UniMM) framework, we recognize critical configurations from both model and data perspectives. We conduct a systematic examination of various model configurations, including positive component matching, continuous regression, prediction horizon, and the number of components. Moreover, our investigation into the data configuration highlights the pivotal role of closed-loop samples in achieving realistic simulations. To extend the benefits of closed-loop samples across a broader range of mixture models, we further address the shortcut learning and off-policy learning issues. Leveraging insights from our exploration, the distinct variants proposed within the UniMM framework, including discrete, anchor-free, and anchor-based models, all achieve state-of-the-art performance on the WOSAC benchmark.

Problem

• Autonomous Vehicles
• Multi-Agent Simulation
• Behavioral Prediction

Innovation

• Hybrid Modeling Framework
• Continuous and GPT-like Models
• Closed-loop Sampling