A Sensorless Reflecting Control for Bilateral Haptic Teleoperation System based on Pneumatic Artificial Muscle Actuators

Abstract

This thesis builds the bilateral haptic teleoperation system (BHTS) driven by pneumatic artificial muscle (PAM) actuators applying studied control approaches. The BHTS enables human users to conduct remote complex tasks in spatial decoupling or inaccessible environments, while can be able to provide the haptic perception of being present for the human operator. However, it is very difficult to ensure a highly accurate performance as well as properties of the BHTS due to incorporation simultaneously between the master-human and slave-environment, the existence of uncertain, nonlinear and unknown terms inside the PAM system dynamics. The strategy of this study begins with developing independent control for each feature of the master and slave system. To start with, the thesis presents several methods to determine both mathematical and experimental models of the haptic bilateral system. Then, a novel integral terminal sliding mode control (ITSMC) scheme is designed as a combination of a nominal control scheme and a discontinuous robust control scheme. Moreover, and time-delay estimator (TDE) technique and an adaptation gain for the ITSMC scheme are not only eliminated uncertainty but also developed to faded the chattering of the control signal and estimate the ideal robust gain. All of them are analyzed via a Lyapunov-based algorithm. The positioning task of both the master/slave systems applies the above-developed controllers and achieve great efficiency. Meantime, for force tracking performance, a fast finite-time nonlinear controller is developed from the fast ITSMC framework with TDE and adaptive switching terms and a friction-free disturbance combined state observer to compensate disturbances. After robot-specific loop implementations are performed, the controller for the bilateral haptic teleoperation control is designed wherein the interactive forces are completely estimated by the upgraded a novel adaptive force observer (AFOB) to external forces without using any force sensor and an advanced state control method is adopted to stabilize the closed-loop system. The robustness, transparency, effectiveness and feasibility of the proposed methods are definitely proven by the theoretical proof and comparative experiments. Especially, to increase the responsiveness to an unknown environment and human behavior, several additional studies in turn for slave-environment interaction and human-mater interaction are based on the learning mechanism. Specifically, a model-free control based on the reinforcement learning (RL) method is established to optimize the desired force from slave-environment interaction. Meanwhile, an integral RL technique is then integrated to adjust the parameter of the prescribed impedance model for human-mater interaction. All proposed algorithms are not only demonstrated by the theoretical proof but also verified in experimental trials.