Inverse Reinforcement Learning for Two-Player Zero-Sum Games

InReinforcement Learning (RL) Chap. 6 , we demonstrated how to solve inverse RL or inverse optimal controlInverse Optimal Control (IOC) problems for linear and nonlinear systemsNonlinear system in a data-driven and model-free manner, assuming no external disturbancesExternal disturbance. In this chapter, we extend our focus to systems that are subject to non-cooperative and adversarial inputsAdversarial input. Such systems are commonly found in various applications, including aircraft, automobiles, electric power systems, economic entities, computer networks, manufacturing, and industrial systems. In control theory, the objective is to find control inputsControl input that counteract disturbances and stabilize these systems. The framework of zero-sum games (Lewis et al. 2012) provides a powerful method to achieve this goal.