Comparative Study of Control Strategies for Under Actuated Manipulator

Abstract

Underactuated mechanical systems are systems with fewer control inputs (actuators) than degrees of freedom. They arise in many applications such as underwater robots, mobile robots and manipulators with passive joints. This thesis considers two-link manipulator with one active joint at shoulder and develops various control strategies for stabilizing the manipulator at the upright position. In nonlinear control, we implemented partial feedback linearization technique for comparison study. We developed sliding mode controller for swinging up the manipulator to the upright position and implemented mono layer sliding mode controller for balancing the manipulator at upright position. In intelligent control, we developed Neuro fuzzy controllers for swinging up and balancing the manipulator to the upright position. In hybrid control, we developed genetic sliding mode controllers for balancing the manipulator at upright position, and tuned the PD gains of feedback linearization technique using genetic algorithm. Results showed that feed back linearization and LQR methods are good for ideal cases when manipulator parameters are precisely known and noise and disturbances are neglected exists. These methods are unable to stabilize the manipulator at the upright position in the presence of certain levels of noise and external disturbance. By tuning the PD gains adaptively using genetic algorithm, feedback linearization tolerates better parameter uncertainties and noise. However, the computational time of Genetic Algorithm is relatively long which prevents the real time implication for short time constant systems. Sliding mode swinging up controller and sliding mode balancing controller are successful techniques for balancing the manipulator at the upright position. They are robust and tolerate noise and external disturbance. The genetic sliding mode controller improves the robustness, reduces the chattering and reduces the hitting time of sliding surface. Neuro fuzzy swinging and balancing controllers are successful in balancing the manipulator at upright position; however, they were sensitive to noise in signals and external disturbances.