Gradient-based Trajectory Optimization with Parallelized Differentiable Traffic Simulation

TL;DR

This paper introduces a scalable, differentiable traffic simulator based on IDM that enables real-time large-scale trajectory optimization, filtering, reconstruction, and prediction, validated on real-world datasets.

Contribution

The authors develop a parallelized, differentiable traffic simulator capable of real-time simulation of up to 2 million vehicles, enabling efficient gradient-based trajectory optimization and analysis.

Findings

Simulates up to 2 million vehicles in real time

Effective noise filtering and trajectory reconstruction

Accurate future trajectory prediction

Abstract

We present a parallelized differentiable traffic simulator based on the Intelligent Driver Model (IDM), a car-following framework that incorporates driver behavior as key variables. Our vehicle simulator efficiently models vehicle motion, generating trajectories that can be supervised to fit real-world data. By leveraging its differentiable nature, IDM parameters are optimized using gradient-based methods. With the capability to simulate up to 2 million vehicles in real time, the system is scalable for large-scale trajectory optimization. We show that we can use the simulator to filter noise in the input trajectories (trajectory filtering), reconstruct dense trajectories from sparse ones (trajectory reconstruction), and predict future trajectories (trajectory prediction), with all generated trajectories adhering to physical laws. We validate our simulator and algorithm on several datasets including NGSIM and Waymo Open Dataset. The code is publicly available at: https://github.com/SonSang/diffidm.