Perturbation Observer-Based Obstacle Detection and Its Avoidance Using Artificial Potential Field in the Unstructured Environment

Abstract

Different methodologies for manipulators have been proposed and applied to robot obstacle detection and avoidance in unstructured environments. These methods include different real-time sensors, observer-based algorithms, and path planning using genetic algorithms. However, sensor design integration is complex and considerably expensive. Moreover, the observer algorithm requires complete system dynamics information, which is difficult to derive. In this regard, genetic algorithms are typically considered slow and difficult to optimize. Accordingly, this study proposes a sensor-less obstacle detection technique using a nonlinear observer (known as sliding perturbation observer (SPO)). Obstacle avoidance is also implemented using a motion planner (known as artificial potential field (APF)). The SPO is a nonlinear observer that only requires the partial position and provides all other states (such as position, velocity) and perturbation (non-linearities and external disturbance). The SPO estimates the external torque at each joint resulting from contact (i.e., collision) with an obstacle. Obstacles are detected and avoided by integrating the SPO and APF. The estimated external torque detects the obstacle location and a repulsive force from the APF is applied to avoid this obstacle. To achieve obstacle avoidance, the sum of all estimated torques must be zero. The proposed technique is applied to a robot manipulator with five degrees of freedom.