A motion planner for a redundant mobile manipulator using the inverse kinematics

Abstract

This thesis is an effort to develop a powerful motion-planning algorithm for the mobile manipulator. The mobile manipulator is expected to work in partially defined or unstructured environments. In global/local approach, joint trajectories are generated for a desired cartesian space path. In an event of failure to obtain practically feasible trajectories, the task cannot be accomplished. For the local path planner, inverse kinematics for a redundant system is used. Inverse kinematics with cost functions and inverse kinematics with priority for tasks are compared, based on the sensitivity to the error in the inputs. Various constraints on the system are considered in the motion planner. Joint displacement limits for the manipulator links are considered as cost function in the inverse kinematics. Obstacle avoidance is also accounted for in the inverse kinematics by seeking to maximize the distance from the obstacle. The system sampling time is adjusted to abide to the joint velocity limits. The motion does not violate the non-holonomic constraint for the mobile base, as point of reference is on the end-effector of a three-link manipulator arm. The system definition of the three-link mobile manipulator is such that the nonholonomic constraint can be easily tackled. In this work, the use of Jacobian matrix elements as gradients is proposed in such a situation. At the point of failure, a deviation in the desired path is obtained and replanner gives a new path that would help in reaching the goal. To calculate deviation, a non-linear optimization problem is formulated and solved by the standard Sequential Quadratic Programming method.