The Improving of BP Algorithmic and Its Application in Robot´s FSM

Finite state machines have the function of reducing the complexity of the robot control system, so it´s introduced into the architecture of the excavating robot. For the sake of embodying artificial intelligence and strategy in the state changing of finite state machines, BP neural network is employed in the finite state machines of the excavating robot. Aiming at the slow convergence speed of the traditional BP network, the weight and transfer function of the traditional BP network is improved, the adjustability of the learning factor is improved, the momentum factor is added, the steep factor is added in the S-function and the accumulated error back propagation arithmetic is adopted. The finite state machines model of the excavating robot is established and the application effects of the improved BP network and other BP network in finite state machines are simulated.