MDG: Masked Denoising Generation for Multi-Agent Behavior Modeling in Traffic Environments

2025-11-21

AI Summary

Existing diffusion and autoregressive models for traffic multi-agent behavioral modeling suffer from inefficiency and poor generalization due to iterative sampling, sequential decoding, or task-specific architectures. This paper proposes Masked Denoising Generation (MDG), a unified framework that reformulates multi-agent behavior modeling as spatiotemporal tensor reconstruction. Its core innovation is a continuous, agent- and timestep-adaptive noise masking mechanism, enabling localized denoising and controllable trajectory generation—eliminating reliance on discrete diffusion time steps and permitting efficient single-pass forward generation. MDG supports integrated modeling across open-loop prediction, closed-loop simulation, motion planning, and conditional generation. Experiments demonstrate state-of-the-art closed-loop performance on Waymo Sim Agents and nuPlan benchmarks, alongside superior efficiency, trajectory consistency, and fine-grained controllability.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Multiagent Systems: Adversarial Agents
• Machine Learning: Deep Generative Models & Autoencoders

Application Category

• 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
• Search and Retrieval-Augmented AI: Agentic search

Abstract

Modeling realistic and interactive multi-agent behavior is critical to autonomous driving and traffic simulation. However, existing diffusion and autoregressive approaches are limited by iterative sampling, sequential decoding, or task-specific designs, which hinder efficiency and reuse. We propose Masked Denoising Generation (MDG), a unified generative framework that reformulates multi-agent behavior modeling as the reconstruction of independently noised spatiotemporal tensors. Instead of relying on diffusion time steps or discrete tokenization, MDG applies continuous, per-agent and per-timestep noise masks that enable localized denoising and controllable trajectory generation in a single or few forward passes. This mask-driven formulation generalizes across open-loop prediction, closed-loop simulation, motion planning, and conditional generation within one model. Trained on large-scale real-world driving datasets, MDG achieves competitive closed-loop performance on the Waymo Sim Agents and nuPlan Planning benchmarks, while providing efficient, consistent, and controllable open-loop multi-agent trajectory generation. These results position MDG as a simple yet versatile paradigm for multi-agent behavior modeling.

Problem

• Modeling realistic multi-agent behavior for autonomous driving
• Overcoming limitations of iterative sampling in existing approaches
• Unifying trajectory generation across prediction and planning tasks

Innovation

• Masked denoising reconstructs spatiotemporal tensors
• Continuous noise masks enable localized trajectory generation
• Unified framework generalizes across prediction and planning