RAPiD: Real-time Deterministic Trajectory Planning via Diffusion Behavior Priors for Safe and Efficient Autonomous Driving

2026-02-07

AI Summary

Diffusion models for autonomous driving trajectory planning struggle to meet real-time and safety requirements due to their reliance on stochastic iterative sampling. To address this, this work proposes the RAPiD framework, which transforms a pretrained diffusion planner into a deterministic policy through score-regularized distillation, thereby eliminating the sampling process entirely. A predictive control–based driver model is introduced as a safety-oriented critic network to provide supervisory signals. This approach achieves, for the first time, efficient extraction of a deterministic policy from a diffusion planner while preserving its multimodal behavior modeling capability. The resulting method attains an 8× speedup in inference, matches the performance of diffusion-based baselines in nuPlan closed-loop simulation, and achieves state-of-the-art generalization on the interPlan benchmark.

Technology Category

• Planning, Routing, and Scheduling: Planning with Language Models
• Humans and AI: Human-Aware Planning and Behavior Prediction
• Computer Vision: Diffusion Models for Vision

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Diffusion-based trajectory planners have demonstrated strong capability for modeling the multimodal nature of human driving behavior, but their reliance on iterative stochastic sampling poses critical challenges for real-time, safety-critical deployment. In this work, we present RAPiD, a deterministic policy extraction framework that distills a pretrained diffusion-based planner into an efficient policy while eliminating diffusion sampling. Using score-regularized policy optimization, we leverage the score function of a pre-trained diffusion planner as a behavior prior to regularize policy learning. To promote safety and passenger comfort, the policy is optimized using a critic trained to imitate a predictive driver controller, providing dense, safety-focused supervision beyond conventional imitation learning. Evaluations demonstrate that RAPiD achieves competitive performance on closed-loop nuPlan scenarios with an 8x speedup over diffusion baselines, while achieving state-of-the-art generalization among learning-based planners on the interPlan benchmark. The official website of this work is: https://github.com/ruturajreddy/RAPiD.

Problem

• trajectory planning
• diffusion models
• real-time
• autonomous driving
• deterministic policy

Innovation

• deterministic policy extraction
• diffusion behavior priors
• score-regularized policy optimization
• real-time trajectory planning
• safety-focused imitation learning