Multiprocessor based control for multiple robots: Software and kinematic programming methodology

Roderic Lawrence Guptill, Purdue University

Abstract

This thesis describes the components of a multiprocessor robot controller capable of communication and kinematic motion coordination of multiple robotic devices. We classify linked mechanisms with one or more controllable axes of motion into the generic group of robotic devices. This classification forms the basis for controlling different devices in a common way, whether they have six axes or just two. This allows a concise method of position specification for robot tasks requiring the coordination of multiple devices. A motion control library, similar in concept to RCCL, is designed and coded for control of multiple robotic devices. The multiple device position specification and device independent trajectory generation are explained. A task level programming system for path specific robot tasks is designed on top of and interfaced to the multiple device motion library. The task domain, path specific tasks, is those robot tasks in which an exact path the manipulator is to follow is specified, such as welding or sealant application. The initial path selection is made through a graphical interface using a pointing device to outline the desired path on a CAD model of the workpiece. The final result of the system is the optimal location to place the part to perform the desired task. Optimality is based on maximizing the manipulability of the manipulator performing the task using a function of the jacobian. User constraints, joint limit constraints, and collision avoidance constraints are used to guide the optimal location selection. The workable task is then executed using calls to the motion control library outlined in chapter two. Finally, a multiprocessor controller design is proposed on which the multiple robotic device library is integrated. The multiprocessor design allows a variable number of processors to be used depending on the computation complexity of the robot task.

Degree

Ph.D.

Advisors

Ahmad, Purdue University.

Subject Area

Electrical engineering

COinS