Intelligent Model-Free Control for Tendon-Driven Continuum Robotic Arms

Continuum manipulator modeling is always associated with structured and unstructured uncertainties. Therefore, model-based control system design for this class of robotic systems will be very challenging. On the other hand, the performance quality of model-free controllers is completely dependent on their hyperparameters; they are also very sensitive to the considered trajectory for the system. As a result, using model-free controllers will require setting parameters for different scenarios. In this research, the design of a model-free controller with comparable performance to model-based control strategies is presented. To this end, various parameters are determined online by the gain adjustment system. The research innovation is to use a supervised machine learning method, fuzzy inference system (FIS), to implement the intelligent gain adaptation system to achieve this goal. The Modified Transpose Jacobian (MTJ) performs well in trajectory tracking due to its approximated feedback linearization tool. In addition, the PID controller structure makes it a locally robust control strategy. An adaptive gain adjustment system can greatly increase algorithm potential and establish the capability to follow trajectories in different work points in the system work space. This research aims to improve the performance of the MTJ model-free control strategy in tracking trajectories starting from arbitrary initial conditions in the system work space. This is achieved by the gain adjustment system design using a fuzzy inference system. Both simulation and experimental results reveal the merits of the proposed controller.