Minimum set of feedback sensors for high performance decentralized cooperative force control of redundant manipulators

Manipulation tasks by humans require a variety of biological sensors, from visual, kinesthetic, contact, pressure, temperature, etc., as well as lots of involved degrees of freedom, conforming a highly redundant biomechanical system. Besides such sensorpsilas network and the unmatched brain power of humans to carry out this task, biomechanical redundancy is a key issue to solve whether we want to transfer this ability on mechanical robots. A much simpler version of a human manipulation task is a cooperative robotic task, which has been mastered by experimental robots at some degree. In this case, the question is whether robots can achieve high performance (fast, accurate and robust tracking without complete knowledge of system dynamics) for this cooperative task using only sensors based on the Newtonian dynamics (encoders, tachometers and force sensors). In this paper, we find a positive answer to this question provided that precise redundancy resolution is introduced, otherwise an intelligent sensor networks is required. Simulation results of representative cooperative tasks employing 7 degrees of freedom manipulators illustrate this concept and further discussions are presented.