A general framework for the analysis and design of the control system with multiple robots handling a common object is presented, based on the decomposition of the combined system dynamics into motion-control and internal force-control subsystems. It is found that, under the infinite-rigidity assumption, the motion-control subsystem is independent of the internal force control, and has the same form and possesses the same properties as those of a single manipulator. This allows for any advanced motion-control law, developed for the control of a single manipulator, to be applied directly to the motion control of the multiple manipulators. In particular, a decentralised adaptive-control law is incorporated to form the motion-control law for the co-operative system. With the definition of the internal force and introduction of a load-sharing criterion, an internal-force-control strategy is analysed and formulated together with an internal-force-estimation algorithm so that the internal-force-control strategy may be implemented without using direct force feedback. The effectiveness of this approach to the multiple-arm co-operation is validated through a set of experimental tests.