Physics-Informed Diffusion Models for Vehicle Speed Trajectory Generation

TL;DR

This paper introduces a physics-informed diffusion model for generating realistic vehicle speed trajectories, improving over traditional methods by combining soft physics constraints with advanced neural architectures.

Contribution

It presents a novel diffusion framework with a dual-channel representation and soft physics constraints, outperforming existing models in generating realistic vehicle micro-trips.

Findings

CSDI model achieves lower Wasserstein distances indicating better distribution matching.

Generated trajectories are indistinguishable from real data in validation tests.

Method proves effective for scalable, realistic driving profile generation for ITS 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.