An approach to uncertainty compensation using a neural network for multi-manipulator system control

An approach to uncertainty compensation using a mul-tilayer feedforward neural network in multi-manipulator system control is proposed. The proposed approach is developed by formulating the dynamics of the multi-manipulator system in the constrained motion framework. The error-backpropagation algorithm is employed for neural network learning. The teaching signal for neural network learning is derived by analyzing the stability of the closed-loop system. It is shown that if the neural network learns to generate the proper compensating signal, then the constrained motion of the multi-manipulator system tracks the desired motion asymptotically; as a consequence , the desired forces can be achieved. Computer simulations are conducted to verify the proposed approach. A multi-manipulator system consists of two or more robots functioning in a coordinated fashion to accomplish a task. An application of this type of robotic system is in the area of fixtureless assembly, where, for instance, two mechanical manipulators hold two separate work pieces together while the pieces are being bonded [15]. Other applications include such tasks as space station assembly, maintenance, and servicing. Significant research has been reported in the literature on the control of multi-manipulator systems. The two main approaches are the so-called masterlslave formulation (e.g. [12, 11) and the hybrid force/position control method (e.g. [$I). Often implicit in these approaches is the assumption that the dynamic parameters of t,he manipulators involved are known precisely. Such a restrictive assumption undermines the practicality of these approaches. To circumvent this difficulty, other control strategies have been proposed to deal with parameter uncertainty associated with the manipulators and to improve the robustness of the control system [lo, 11, 5 , 171. Recently (artificial) neural networks have been employed in the area of robotic control. Application of neu-ral networks to free motion control [16, 7, 31 and contact task control [6, 41 have been reported. An approach using a neural network for uncertainty compensation in the control of multi-manipulator system is proposed in [18], where the neural network is used in conjunction with a hybrid force/position control scheme. In this paper, we proposed an approach for uncertainty compensation using a neural network in multi-manipulator system control. This proposed approach is formulated in the framework of constrained motion. By formulating the multi-manipulator dynamics within the constrained motion framework (as in [15]), the resulting dynamic equations of motion are expressed in the most natural form in a set of generalized coordinates, thus leading to a simplified …