Neural network techniques for scheduling and routing problems in multihop radio networks

The authors address the problem of link activation or scheduling in multihop packet radio networks, a contention-free form of channel access that is appropriate for many military communication applications. This problem, in almost all of its forms, is a combinatorial-optimization problem of high complexity. They approach this problem by the use of a Hopfield neural network model in which the method of Lagrange multipliers is used to dynamically vary the values of the coefficients used in the connection weights. Extensive software simulation results demonstrate the effectiveness of this approach in producing schedules of optimal length. Issues associated with the extension of this approach to the joint routing/scheduling problem are discussed.<>