Fixed-time Recurrent Neural Networks for Time-variant Matrix Computing and Its Application in Repeatable Motion Planning
Fixed-time recurrent neural network(RNN)models with logarithmic settling time are proposed for solving time-variant neural computing problems.Two novel RNN models are designed and analyzed in detail,deriving the explicit expressions of set-tling time functions and providing the upper bounds of the settling times under any initial condition.Compared with the existing RNN models with fixed-time convergence,the two novel models with logarithmic settling time have a smaller upper bound on the settling time and faster convergence speeds.Taking into account initial conditions located within a region with a definite finite ra-dius,the settling time functions of the RNN models with logarithmic settling time are given,and the upper bounds on the settling time functions in the semi-global sense are derived.Modified RNN models adopt the inverse of the bound to ensure that the semi-global predefined time converges to the exact solution,and its prescribed time is an adjustable parameter.Simulation results of the proposed RNN model for solving time-variant Lyapunov and Sylvester equations are given.The proposed RNNs are applied to the repetitive motion planning of a redundant manipulator with initial errors,and numerical results are presented to verify the effec-tiveness of the proposed RNN models.
万方数据
维普
