A B S T R A C TThe IRIS Facility is a modular, reconfigurable and expandable robot system to be used for experiments in grasping, manipulation and force control. The baseline layout of the Facility will have two manipulators with four rotary joints each. Each manipulator can be easily disassembled and reassembled to assume a multitude of configurations. Each joint is driven by d.c. brushless motors coupled with harmonic cup drives and instrumented with position, and torque sensors. A six d.0.J forcdtorque sensor is mounted at the tip link. Additional manipulators with different joint layouts will be added in the future.The real-time controller of the IRIS Faciliry has also been designed to be modular and expandable. It is based on a nodal architecture with a PC-486 host and an AMD29050 co-processor as the CPU of each secondary node. Each node is capable of controlling 8 joints at I kHz while executing over lo00 FP (Floating Point) operations per joint in each sampling interval. This paper describes the design of the IRIS Facility and its functional capabilities. I n addition, the rationale behind the major design decisions is given.

The motion of coordinated robots manipulating an object con strained by the contact with an environment, in the presence of uncertainty in the dynamic models, represents an important class of control problems. In this article an approach to motion and force control of such multiple coordinated robots, based on an adaptive scheme, is proposed. Three major variables are regulated simultaneously: the object motion, the contact force between the object and the environment, and the internal force. The adaptive controller estimates the unknown parameters of the coordinated robots and object in terms of certain error equations. The convergence of the motion, contact, and inter nal force errors is analyzed using Lyapunov stability theory. The proposed controller is also experimentally evaluated on two coordinated AdeptOne robots. In comparison with fixed parameter controllers, the experimental results show that the proposed controller improves the object motion, contact, and internal force tracking accuracy for a class of robotic systems with uncertain knowledge of the dynamic model.

This paper describes the computer architecture for the IRIS Facility, currently under development at the Robotics and Automation Laboratory at the Universily of Toronto. IRIS Facility is a versatile, multi-degree-offreedom reconfigurable and eqmdable setup for research in grasping and manipulation. The proposed computer architecture involves powerful I O MFLOPS RISC processor nodes and a fast point-to-point communication network Each node includes a host coquter and parallel U0 modules as well. The nodes are controlled by nearly zero-overhead customized real-time OS krnels responsible for task scheduling, communication and user-interface.