TRACS: An Experimental Multiagent Robotic System

TRACS (Two Robotic Arm Coordination System), developed at the GRASP Laboratory at University of Pennsylvania, is an experimental system for studying dynamically coordinated control of multiple robotic manipulators. The systems is used to investigate such issues as the design of controller architectures, development of real-time control and coordination programming environments, integration of sensory devices, and implementation of dynamic coordination algorithms. The system consists two PUMA 250 robot arms and custom-made end effectors for manipulation and grasping. The controller is based an IBM PC/AT for its simplicity in I/O interface, ease of real-time programming, and availability of low-cost supporting devices. The Intel 286 in the PC is aided by a high speed AMD 29000 based floating point processor board. They are pipelined in such a way that the AMD 29000 processor performs real-time computations and the Intel 286 carries out I/O operations. The system is capable of implementing dynamic coordinated control of the two manipulators at 200 Hz. TRACS utilizes a C library called MO to provide the real-time programming environment. An effort has been made to separate hardware-dependent code from hardware-independent code. As such, MO is used in the laboratory to control different robots on different operating systems (MS-DOS and Unix) with minimal changes in hardware-dependent code such as reading encoders and setting joint torques. TRACS utilizes all off-the-shelf hardware components. Further, the adoption of MS-DOS instead of Unix or Unix-based real-time operating systems makes the real-time programming simple and minimizes the interrupt latencies. The feasibility of the system is demonstrated by a series of experiments of grasping and manipulating common objects by two manipulators. Comments University of Pennsylvania Department of Computer and Information Science Technical Report No. MSCIS-90-60. This technical report is available at ScholarlyCommons: http://repository.upenn.edu/cis_reports/573 TRACS: An Experimental Mult iagent Robotic System MS-CIS-90-60 GRASP LAB 230 Xiaoping Yun Eric Paljug Ruzena Bajcsy Department of Computer and Informat ion Science School of Engineering and Applied Science University of Pennsylvania Philadelphia, PA 19104