A New Gradient Annealing Algorithm (GAA) and its Applications in Path Planning of Mobile Robot

Abstract

To deal with the problem that the convergence rate of simulated annealing algorithm (SAA) is very slow, a new hybrid optimal algorithm, gradient annealing algorithm (GAA), combined by SAA and GM, based on analysis of gradient method (GM) and SAA, is proposed. And it is successfully applied to the path planning of the neural network of mobile robot. First, GAA uses quickness searching of GM to obtain a local minimum. Second, by utilizing the abilities of global searching of SAA, it escapes from trapping this local minimum. At last the global minimum is achieved through iterative computation. So the convergence rate being improved. The simulation experiments demonstrate that the computation of GAA is simple, the convergence rate is fast and the robustness of initial value is good also.