Design and evaluation of learning human control algorithm for a Wheeled Inverted Pendulum

Human control strategy has been considered as a robust control method in mastering the complex and dynamic skill. In this paper, we present the problem of how human control strategy can be represented as a learning based approach and how a human strategy controller can be used in controlling dynamically stable while statically unstable, Wheeled Inverted Pendulum (WIP). The controller is designed by learning human strategy using Support Vector Regression (SVR). Since most human control strategy controllers are usually empirically rather than analytically, there exists a profound need to evaluate the performance of the human strategy controller. We then propose an effective approach for stability analysis of the human strategy controller based on the discrete-time system Lyapunov theory and Chebyshev points based estimation approach. The experimental results validate our proposed scheme for human controller design and stability analysis.