This paper presents a discrete-time recurrent neural network, with a fixed
step parameter, for solving the shortest path problem. The proposed discret
e-time recurrent neural network with a simple architecture is proven to be
globally convergent to exact optimal solutions and is suitable for hardware
implementation. Furthermore, an improved network with a larger step size i
ndependent of the problem size is proposed to increase its convergence rate
. The performance and operating characteristics of the proposed neural netw
ork are demonstrated by means of simulation results.