Physics-Informed Diffusion Models for Vehicle Speed Trajectory Generation

2026-02-04

AI Summary

This work addresses the limitations of traditional Markov chain approaches in synthesizing vehicle speed trajectories, which suffer from discretization artifacts and insufficient expressiveness. The authors propose a diffusion model framework that incorporates physical priors, introducing soft physical constraints into conditional microtrip generation for the first time. By employing a dual-channel representation of speed and acceleration, the method avoids optimization conflicts caused by hard constraints. The model integrates a 1D U-Net with a Transformer-based Conditional Score-based Diffusion Imputation (CSDI) architecture and is trained on 6,367 GPS-derived microtrips. Experimental results demonstrate that CSDI significantly outperforms baseline methods in matching both speed and acceleration distributions, achieving Wasserstein distances of 0.30 and 0.026, respectively. The generated trajectories exhibit high realism (discriminator score: 0.49) and effectively support downstream vehicle energy consumption estimation tasks.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Planning, Routing, and Scheduling: Model-Based Reasoning
• Search and Optimization: Sampling/Simulation-based Search

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• User Modeling, Personalization and Recommendation: User privacy protection in personalized systems
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Synthetic vehicle speed trajectory generation is essential for evaluating vehicle control algorithms and connected vehicle technologies. Traditional Markov chain approaches suffer from discretization artifacts and limited expressiveness. This paper proposes a physics-informed diffusion framework for conditional micro-trip synthesis, combining a dual-channel speed-acceleration representation with soft physics constraints that resolve optimization conflicts inherent to hard-constraint formulations. We compare a 1D U-Net architecture against a transformer-based Conditional Score-based Diffusion Imputation (CSDI) model using 6,367 GPS-derived micro-trips. CSDI achieves superior distribution matching (Wasserstein distance 0.30 for speed, 0.026 for acceleration), strong indistinguishability from real data (discriminative score 0.49), and validated utility for downstream energy assessment tasks. The methodology enables scalable generation of realistic driving profiles for intelligent transportation systems (ITS) applications without costly field data collection.

Problem

• vehicle speed trajectory generation
• synthetic data
• Markov chain limitations
• realistic driving profiles
• intelligent transportation systems

Innovation

• Physics-Informed Diffusion
• Conditional Micro-Trip Synthesis
• Soft Physics Constraints
• Score-Based Generative Modeling
• Vehicle Speed Trajectory Generation