Abstract: The idea and formal description of channel routing are introduced, and a mean field annealing neural network approach is adopted in the problem. In the algorithm, the total wire-length is shortest and the number of channel is minimized, which is the optimization target based on the constraints of horizontal and vertical wiring. The problem is mapped to the neural network, and the energy function of the problem is established, then the iteration procedure is executed with the mean annealing equation. Normalization of neuron proves that one wire-net can only be assigned to one channel. Its testing result is satisfactory.