| Using multiple robots to perform cooperative tasks such as material transport or parts assembly is becoming increasingly common as the costs of robotic hardware, processing power and software continue to decrease. In such scenarios, the multiple-robot approach provides the redundancy, reliability, speed, payload force regulation and cost-effectiveness that a single robot cannot provide.; There are many areas such as underwater and space exploration, service robotics, etc. that would benefit from multi-robot manipulation. Multiple-robots are most frequently used when the task requirement is incompatible with a single robot. For example, a heavy load that is beyond the capability of a single manipulator may be handled by multiple smaller robots.; Motivated by these observations, this thesis addresses the multiple-robot collaborative manipulation problem, focusing on the motion and force control laws for decentralized and distributed systems. Inspired by the behavior of ants, and based on the move/squeeze decomposition, a hybrid motion/force control is proposed for a planar manipulation of an object by point mass robots. The motion/squeeze decomposition makes it possible to consider the motion and force control as two, one-way coupled, problems. In this way, first the motion control is designed without taking into account the force control. Then, the force control is addressed considering the motion control as a perturbation.; In a new stability result, using a modified energy-based Lyapunov function, the decentralized motion control is shown to be globally asymptotical---locally exponential---stable. The convergence of the motion control is characterized for two cases: if the position set points are kinematically feasible, the position errors of all robots converge to zero. If the position set points are kinematically infeasible, the position errors of the robots converge to the squeeze subspace. This result is used to generate a squeeze force set point for the decentralized and distributed force controllers.; Based on the motion control exponential stability result, for a load initially at rest, the decentralized force control is found to be asymptotically convergent. For the distributed force control, a robustness bound for the maximum communication delay is obtained. Simulation and experimental results have been included to illustrate the results. |
