Risk-Controllable Multi-View Diffusion for Driving Scenario Generation

2026-03-12

AI Summary

Existing methods struggle to generate geometrically consistent long-tail, high-risk driving scenarios and typically treat risk as a post-hoc label, lacking controllable synthesis of multi-view dynamic scenes. This work proposes RiskMV-DPO, the first framework enabling risk-controllable multi-view driving scene generation. It leverages physics-driven risk modeling with high-risk trajectories as geometric anchors and integrates diffusion models to produce spatiotemporally coherent and geometrically precise video sequences. The approach introduces a novel geometry-appearance alignment module and a region-aware direct preference optimization (RA-DPO) strategy, shifting world models from passive prediction toward active risk synthesis. Evaluated on nuScenes, the method significantly improves 3D detection mAP (from 18.17 to 30.50) and reduces FID to 15.70, successfully generating diverse, high-fidelity long-tail risk scenarios.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Humans and AI: Human-Aware Planning and Behavior Prediction
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Generating safety-critical driving scenarios is crucial for evaluating and improving autonomous driving systems, but long-tail risky situations are rarely observed in real-world data and difficult to specify through manual scenario design. Existing generative approaches typically treat risk as an after-the-fact label and struggle to maintain geometric consistency in multi-view driving scenes. We present RiskMV-DPO, a general and systematic pipeline for physically-informed, risk-controllable multi-view scenario generation. By integrating target risk levels with physically-grounded risk modeling, we autonomously synthesize diverse and high-stakes dynamic trajectories that serve as explicit geometric anchors for a diffusion-based video generator. To ensure spatial-temporal coherence and geometric fidelity, we introduce a geometry-appearance alignment module and a region-aware direct preference optimization (RA-DPO) strategy with motion-aware masking to focus learning on localized dynamic regions.Experiments on the nuScenes dataset show that RiskMV-DPO can freely generate a wide spectrum of diverse long-tail scenarios while maintaining state-of-the-art visual quality, improving 3D detection mAP from 18.17 to 30.50 and reducing FID to 15.70. Our work shifts the role of world models from passive environment prediction to proactive, risk-controllable synthesis, providing a scalable toolchain for the safety-oriented development of embodied intelligence.

Problem

• risk-controllable generation
• multi-view consistency
• driving scenario generation
• long-tail scenarios
• geometric fidelity

Innovation

• risk-controllable generation
• multi-view diffusion
• geometry-appearance alignment
• region-aware DPO
• driving scenario synthesis