A digital neural network for the traffic control problem on crossbar switch networks

The traffic control problem on crossbar switch networks is represented by a parameter-free energy function. The proposed neural network is applied to update the state of the energy function until a stable state is reached. Within O(n) iteration steps, where n is the size of an n*n network, the energy function reaches a stable state which corresponds to a feasible solution of the traffic control problem. Simulation results show that this neural network also generates either optimal or near-optimal solutions.<
>