TRAJECTORY FOLLOWING OF ROBOTIC MANIPULATOR

Abstract

Trajectory following is a typical application of robotic manipulators. It is desired that the robot has the potential to adjust its motion in a changing environment. This can be achieved based on on-line programming with the aid of extra sensors. The sensors provide real-time information of the environment. The real-time information together with the desired trajectory are used for the trajectory planning. Both non-redundant manipulators and redundant manipulators can be used as platforms to achieve trajectory following task. For more complex tasks, redundant manipulators are more flexible than non-redundant manipulators. This thesis focuses on the trajectory planning on both non-redundant and redundant manipulators. A typical contour-following task is implemented on the PUMA560, a standard nonredundant robot. Contour-following is one kind of trajectory following based on online programming. It is required to move the end-effector along a specified direction while keeping the end-effector parallel with a surface at a desired offset distance from the surface. Three laser sensors are mounted onto the end-effector to obtain the environment information. An algorithm is developed to compute the homogeneous transformation matrix of the end-effector at every motion step. Two control strategies are used to implement the contour-following on the PUMA560 robot. One is the independent joint space control. Another is model-based Cartesian space control. Experiments are done to accomplish the contour-following task along a straight-line trajectory and a zig-zag trajectory The extra degrees of freedom introduced in redundant manipulators enhance the manipulators with the ability to achieve additional tasks beside following a desired trajectory. Normally, these additional tasks are specified by some performance criteria. To fully take advantage of the ability, it is desired to resolve the redundancy. By incorporating the desired trajectory and extra performance criteria, the redundancy can be resolved in joint position level, joint velocity level or joint acceleration level. Obstacle avoidance problem is a typical application of redundant manipulators. A redundant manipulator is required to follow a desired trajectory while avoiding the collision between links and obstacles. In literature, this problem is resolved in joint velocity level based on the pseudoinverse of the Jacobian matrix. Motion cyclicness is a desired joint position level property of redundant manipulators. The final joint configuration is required to converge to the initial joint configuration for a closed trajectory. In the case of obstacle avoidance problem, the existence of a closed-even trajectory is the necessary condition to achieve motion cyclicness. A definition of closed-even trajectory is presented. A new approach based on distance optimization is proposed to resolve the redundancy in joint position level. It is formulated as a non-linear optimization problem. The approach enhances the obstacle avoidance capability of redundant manipulators. Since the approach resolves the redundancy in joint position level, motion cyclicness can be easily included and achieved. One requirement behind the proposed approach is the ability to compute the shortest distance between two convex objects. This can be achieved by using the distance computation routines. Simulations are performed in 2-D and 3-D spaces. In 2-D case, the simulation is based on a planar redundant manipulator. In 3-D case, the simulation is based on a 4-DOFs redundant manipulator derived from the PUMA560 geometry. The simulation results demonstrate the usefulness of the approach.