Semi-global Convergence and Finite-Sample Error Bounds for Model Predictive Path Integral Control
Christian Fiedler, Giacomo Borghi
Abstract
We investigate the global optimisation behaviour of Model Predictive Path Integral Control (MPPI) in the low-temperature setting. For objective functions with a unique global minimum and under weak additional assumptions, allowing non-convexity and non-smoothness, we establish semi-global convergence of the mean-field or population level MPPI iteration to this global optimum. We also establish finite-sample results that show that with high probability the MPPI iteration comes close to the global optimum and provide sample complexity results. Under slightly stronger assumptions, we make these results more quantitative and provide concrete bounds on the inverse temperature and number of samples. Finally, we illustrate the results on the classic LQR problem.
Create a lesson
Related papers
Optimal Scheduling of Multi-Site AI Data Centers in Day-Ahead Markets
Xinyang Zhou, Guido Cavraro
Minimization of free energy in a Poisson--Boltzmann problem
Shahid Ali, Klemen Bohinc, Angeles Martinez et al.
Lyapunov Theory for Output-to-State Stability of Nonlinear Infinite-Dimensional Systems
Qiaoling Chen, Andrii Mironchenko, Fabian Wirth
When Suspicion Adds Up: A Stochastic Framework for Optimizing Honeypot-Based Insider-Threat Detection
Saharnaz Mehrani, Lujia Zhan, Mohsen Emadikhiav et al.
Hamiltonian Potential Optimization for Partially Observed Stochastic Optimal Control
Christian Bayer, Saifeddine Ben Naamia, Erik von Schwerin et al.
Paper2Lean: Agentic Autoformalization of Mathematical Research Papers at Scale
Yifan Bai, Yantao Li, Jian Yu et al.