GoSafeOpt: Scalable Safe Exploration for Global Optimization of Dynamical Systems
This addresses the challenge of safe exploration in robotics and control systems, enabling global optimization without hardware damage, though it builds incrementally on prior work like GoSafe.
The paper tackles the problem of safely learning globally optimal control policies for high-dimensional dynamical systems without failures, and demonstrates that GoSafeOpt outperforms existing model-free safe learning methods on a robot arm.
Learning optimal control policies directly on physical systems is challenging since even a single failure can lead to costly hardware damage. Most existing model-free learning methods that guarantee safety, i.e., no failures, during exploration are limited to local optima. A notable exception is the GoSafe algorithm, which, unfortunately, cannot handle high-dimensional systems and hence cannot be applied to most real-world dynamical systems. This work proposes GoSafeOpt as the first algorithm that can safely discover globally optimal policies for high-dimensional systems while giving safety and optimality guarantees. We demonstrate the superiority of GoSafeOpt over competing model-free safe learning methods on a robot arm that would be prohibitive for GoSafe.