Safe Online Control-Informed Learning

TL;DR

This paper introduces a unified framework for safe, real-time learning in autonomous systems that combines control, estimation, and safety constraints, ensuring robustness and safety during online operation.

Contribution

It presents a novel framework integrating optimal control, parameter estimation, and safety constraints with real-time updates and safety guarantees, applicable to safety-critical systems.

Findings

Successful demonstration on cart-pole system

Effective safety constraint enforcement during learning

Robust parameter estimation under uncertainty

Abstract

This paper proposes a Safe Online Control-Informed Learning framework for safety-critical autonomous systems. The framework unifies optimal control, parameter estimation, and safety constraints into an online learning process. It employs an extended Kalman filter to incrementally update system parameters in real time, enabling robust and data-efficient adaptation under uncertainty. A softplus barrier function enforces constraint satisfaction during learning and control while eliminating the dependence on high-quality initial guesses. Theoretical analysis establishes convergence and safety guarantees, and the framework's effectiveness is demonstrated on cart-pole and robot-arm systems.