Controller Fusion Based on the Velocity Vector for Redundant Manipulator

Redundant manipulator control involves the mapping of error measured in Cartesian space (where the task is defined) to joint space (where the actuators are controlled). In classical algorithms, the mapping is made through Jacobian pseudo-inverse matrix that has a high computational cost. This work proposes a new approach for control of redundant manipulators based on controller fusion techniques. In this approach the control task is divided in subtasks. For example, a subtask can be tool positioning in a single Cartesian dimension, or obstacle avoidance, or singularity avoidance. A controller is implemented for each subtask, and these simple controllers are called subtask controllers. The resultant signals are fused to generate the control signal to be applied to the robotic arm. The fusion system analyzes the resultant signals of the subtask controllers associating a weight to each one. These weights depend of the control task. When the control task is trajectory following, a measure based on the trajectory variation is necessary. In this paper, the measure is obtained from the velocity of the reference trajectory. Preliminary tests showing the viability of subtask controllers and simulation results of the fusion algorithms are presented