Diffusion-Based Planning for Autonomous Driving with Flexible Guidance

2025-01-26

AI Summary

Human-like driving planning in complex, open-road environments requires simultaneous multi-objective optimization, accurate modeling of intricate driving behaviors, and adaptive style generalization—challenging for existing end-to-end approaches. Method: This paper proposes the first diffusion-based end-to-end closed-loop planning framework, integrating a Transformer architecture with gradient estimation of trajectory score functions and a flexible classifier-free guidance mechanism. It enables joint prediction-planning learning and vehicle-to-vehicle cooperative modeling, eliminating rule-based post-processing while inherently ensuring trajectory safety and style adaptability. Contributions/Results: Evaluated on nuPlan and a proprietary 200-hour delivery-vehicle dataset, our method achieves state-of-the-art closed-loop performance, demonstrating significant improvements in cross-style generalization and operational safety without compromising real-time feasibility.

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

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Achieving human-like driving behaviors in complex open-world environments is a critical challenge in autonomous driving. Contemporary learning-based planning approaches such as imitation learning methods often struggle to balance competing objectives and lack of safety assurance,due to limited adaptability and inadequacy in learning complex multi-modal behaviors commonly exhibited in human planning, not to mention their strong reliance on the fallback strategy with predefined rules. We propose a novel transformer-based Diffusion Planner for closed-loop planning, which can effectively model multi-modal driving behavior and ensure trajectory quality without any rule-based refinement. Our model supports joint modeling of both prediction and planning tasks under the same architecture, enabling cooperative behaviors between vehicles. Moreover, by learning the gradient of the trajectory score function and employing a flexible classifier guidance mechanism, Diffusion Planner effectively achieves safe and adaptable planning behaviors. Evaluations on the large-scale real-world autonomous planning benchmark nuPlan and our newly collected 200-hour delivery-vehicle driving dataset demonstrate that Diffusion Planner achieves state-of-the-art closed-loop performance with robust transferability in diverse driving styles.

Problem

• Autonomous Vehicles
• Multi-object Tracking
• Adaptive Driving Styles

Innovation

• Transformer-based Planner
• Closed-loop Planning
• Adaptive Driving Styles