Task-driven control of multi-arm systems

Abstract

Task-driven robot control schemes follow changing task conditions, have servos in task space and have a natural relation to the task. Therefore, task-driven schemes can help in achieving 'intelligent' robot arm behavior. In this paper a generic task-driven servo scheme is discussed for the problem of object handling by a multi-arm robotic system. The primary goal is to make the object exhibit a prescribed behavior in the presence of an environment. Four classes of task-level entities are identified: Position, Environmental Interaction Force, Internal Force and 'Zero Dynamics' or Auxiliary Coordinates. Sensed task-level errors directly drive the system through a nonlinear feedback control law. Redundancies that exist at object and robot arm levels are identified. Novel strategies to resolve redundancy in joint motion are suggested. Full use of the number of available inputs is made, and servo loops equal to the total number of inputs, control the system as a whole. Numerical simulations for a dual-arm situation illustrate the validity of the approach. Furthermore, a novel technique based on singular value decomposition is described here to analyze practical situations in object handling. Results of experiments done on a dual six-degree-of-freedom arm arrangement, with force sensing, are reported.