Continuous-time optimal control for trajectory planning under uncertainty

TL;DR

This paper develops a continuous-time optimal control framework for autonomous vehicle trajectory planning under uncertainty, demonstrating improved robustness, faster computation, and better constraint adherence compared to discrete-time models.

Contribution

It extends previous discrete-time stochastic models to continuous time, enhancing robustness and computational speed for real-time trajectory planning in diverse driving scenarios.

Findings

Continuous-time model captures uncertainty more effectively.

Faster computation with dynamic solvers.

More robust to various driving scenarios.

Abstract

This paper presents a continuous-time optimal control framework for the generation of reference trajectories in driving scenarios with uncertainty. A previous work presented a discrete-time stochastic generator for autonomous vehicles; those results are extended to continuous time to ensure the robustness of the generator in a real-time setting. We show that the stochastic model in continuous time can capture the uncertainty of information by producing better results, limiting the risk of violating the problem's constraints compared to a discrete approach. Dynamic solvers provide faster computation and the continuous-time model is more robust to a wider variety of driving scenarios than the discrete-time model, as it can handle further time horizons, which allows trajectory planning outside the framework of urban driving scenarios.