ROBOTICS PUBLICATIONS

A review of the authors' developments in Reflex Control is presented, with a focus on on-line protection against higher-level software errors. Collision-avoidance reflexes are described, from protection against self-collisions and collisions with static obstacles, to collision prevention in the context of multi-arm coordination. Theory of operation and experimental results are described. It is shown how reflex control can enhance the survivability of complex systems without degrading performance. Further, it is illustrated how the existence of low-level reflex behaviors can simplify the design of higher-level controls and planners. Reflex control is described as a valuable component of robust, autonomous systems.

Using reflex control, we have achieved guaranteed obstacle avoidance taking into account robot dynamics. Reflex control is a fast, on-line and fail-safe obstacle avoidance method that protects the system from erroneous higher-level commands. The reflex controller inspects nearby configuration space and does not approve commands that would result in collision with an obstacle, even under high-speed motions. A new generalized approach to implement reflex control is described.

A method is presented for computation of obstacle boundaries in configuration space with particular emphasis on time efficiency. Properties of configuration space transformations are presented in general, and these properties are invoked in algorithms which result in highly efficient computations. The approach depends on the definition of a set of primitives which are themselves efficiently transformed and may be combined logically to construct more complex transformations. The concepts are illustrated with both computed and empirically obtained configuration-space obstacles in 2-D and in 3-D. Performance data for real-time transformations is reported.

Experiments in automatic collision avoidance for robots using acceleration-based reflex control are described. Acceleration-based reflex control responds to acceleration commands from higher levels, as opposed to prior reflex control approaches based on position commands from higher levels. Reflex control is exerted in configuration space and assumes either inherent dynamic decoupling of the robot links or feedback decoupling. Implementation of acceleration-based reflex control is described for 3-D collision avoidance using a General Electric GP132 robot. Collision avoidance with respect to stationary obstacles is guaranteed. In addition, performance measurements illustrate how the reflexes introduce no significant distortion to higher-level controllers when reflex action is not required.

Mathematical properties of configuration space are presented, and algorithms invoking those properties for efficient computation of obstacles in configuration space are described. Simple elements in Cartesian space are identified which can be transformed into configuration space rapidly. Transformations of complex workspace shapes into configuration space are described in terms of multiple transformations of such simpler primitives. Computational considerations and examples are presented for the first three degrees of freedom of an industrial robot.

This paper describes a systems approach to improving the performance of a telerobotic testbed. The testbed is the Force-Reflecting Interfaces to Telemanipulators Testing System (FITTS), at the Harry G. Armstrong Aerospace Medical Research Laboratory. First, the testbed hardware is described, along with an overview of the system's communications paths. Next, the previously-determined forward kinematics for the MBA exoskeleton are outlined. Then the optimization of these kinematic equations is given. The paper also details efficient forward and inverse kinematic solutions for the Merlin industrial robots, using the method of wrist partitioning. The utility of these solutions in toto is that the interfacing of the two systems, given sufficient communications bandwidth between the MBA exoskeleton and the control computer, can now achieve the 4 ms compute time attainable by the Merlin hardware. This optimization puts the FITTS hardware at a milestone stage of completion, as the system is now capable of operating at its peak as a unilateral telerobotic testbed. Finally, future steps to extend FITTS for force reflection research are specified.

Two projects aimed at improving human sensory feedback capabilities of existing systems were undertaken. The first involved modeling and control of the Force Reflecting Elbow Bilateral Exoskeleton (FREBiE) system; the second improved control of Merlin industrial robots via the MBA exoskeleton (FITTS system). The main concentration with the FREBiE was on modeling and then compensating for its significant nonlinear dynamics, which were found to be gravity and Coulomb friction. Accurate linearization of the FREBiE system resulted in drastically improved performance. It permits the addition and testing of novel control laws and allows meaningful data collection. The FITTS project concentrated on formulating forward and inverse kinematic models for the Merlin robots and forward kinematics for the MBA exoskeleton, allowing ``filtering'' of position and orientation parameters, and improving controller bandwidth. It culminated in completion of a teleoperated peg-in-the-hole insertion task with the Merlin robot being commanded via the MBA exoskeleton. Providing this capability allows comparisons to be made with the data for a battery of such tests already collected for unencumbered humans and humans encumbered by the MBA exoskeleton. It also establishes a basis for implementing novel control architectures to improve human-robot interaction and performance. Finally, it may testify to the need for different types of sensory feedback and provide a reference point for such experiments.