A Virtual Security Net and Soft Motion Controller for a 7-DOF Redundant Cooperative Robotic Manipulator

Abstract

This thesis presents the study and implementation of a virtual security net for a cooperative robotic manipulator. The input to the security net includes a goal- position and orientation for the end-effector paired with the desired approach direction. The security net ensures that the corresponding output is a safe trajectory from start pose to goal pose that the end-effector successfully tracks. Moreover, the security net consists of a global path planning and obstacle avoidance method, a trajectory generation technique based on waypoints, and a reliable soft motion controller that possesses considerable tuning capabilities. Comprehensive literature studies have been made in several relevant fields, and this has formed a basis for the development process of the security net. In brief details, the path planning algorithm is based on a distance transform (DT) on a three-dimensional workspace to avoid obstacles and control the speed of the robot. Moreover, the graph search algorithm $A^*$ is used in conjunction with the DT to calculate an optimal path for the robot in cluttered environments. The trajectory generation technique creates cubic a B-spline which satisfies $C^2$ continuity, which a velocity guidance logic is built on. Finally, the soft motion controller is defined as a Quadratic Programming problem in order to conveniently minimize tracking errors and induce penalties on undesirable behaviors, as well as providing a reliable and simple way to handle robot constraints.