Robust adaptive controller based on evolving linear model applied to a Ball-Handling mechanism

Abstract

The increasing complexity and permanent growth of real-world robotics formidable challenges demand that most control systems be intelligently adaptive to the parameters and structures of dynamics. This paper, therefore, discusses an extended sliding mode controller that is based on an evolving linear model (ELM) designed and implemented as a systematic approach to tackling the arms target angle tracking problem in the ball-handling system of a robot. Without any prior knowledge about the dynamics of the system other than its highest possible order, the dynamic orders and relative degrees of the system are practically derived. A novel online linearization technique based on the recursive least squares (RLS) method which keeps the output error of estimation in a relatively small bound is applied to identify the plant and to derive an adaptive-linear-regression (ALR) model of the system. Subsequently, having a model in which the number of constructing independent regressors varies over time, an extended sliding mode control strategy, established upon Lyapunov theory, is applied to the online-identifying ELM of the ball-handling system. In order to quantify the effectiveness of the proposed methodology, comparative analysis of the proposed strategy with well-established linear quadratic regulator (LQR) design and other suggested work on this topic, on the robustness of controllers, are performed in simulations. Ultimately, multifarious practical scenarios were designed, performed, and validated for the handling mechanism. The results clearly demonstrate the benefits and effectiveness of the design approaches in practice.