Abhishek Kumar
Reinforcement learning, stability analysis, intelligent control, Lyapunov stability, survey
Reinforcement learning (RL) is one of the most emerging domains of artificial intelligence. It is widely used in almost all sort of applications, including medical field, stock market, forecasting, and engineering field. One of the most effective uses of RL is in the control engineering domain owing to its learning by trial, for example, in building autonomous system like autonomous vehicle and robotics. In this paper, we have focused on the applications of RL in various control engineering problems. Stability of controller (or agent) using RL paradigm is a very crucial task due to exploration–exploitation policy used by any RL. Also, the unavailability of exact model of system or environment may lead to unsafe behaviour of the agent. Therefore, this paper focuses mainly on the stability aspect in RL-based controller. Many concepts are used to study and analyse the stability of a system viz Lyapunov method and Barrier function. This paper surveys the detailed application of these well-established stability certifier methods in various model-free and model-based RL framework.
Important Links:
Go Back
