APPLICATION OF SELF-ORGANIZING NEURAL NETWORKS IN ROBOT TRACKING CONTROL

The use of a self-organising neural network as a feedforward compensat or for robot tracking control applications is proposed. The topology o f the input space is adaptively mapped onto a set of neurons where eac h neuron represents a discrete cell in the input domain. Within each c ell, a linear mapping is established between the input and output spac e. The training of such a network involves training of a weight vector that represents the topology of the input space and weight vectors (a ction space weights) that linearly code an input pattern to action spa ce. In the first phase of network training, a 'neural-gas' algorithm i s employed for learning the topology of the input space while weight v ectors representing control action space is learned by backpropagating feedback control action. During this phase of learning, the weights a ssociated with neurons in the neighbourhood of winning neurons are als o updated. In the second stage, a recursive least squares based estima tion scheme is applied to fine tune the action space weights associate d with winning neurons only, without disturbing the input topology map learned in the first phase. The proposed scheme has been compared wit h multilayered network (MLN) and radial basis function network (RBFN) based inverse dynamics learning schemes. Simulation results show that the proposed scheme has better generalisation capability than both MLN and RBFN.