A neural network solution to the multicast packet radio transmission problem

This work describes a practical technique for the computation of optimum or near optimum paths from a single source to multiple destinations in a packet radio network (PRN) environment. The problem is common and is usually solved by making copies of the packet, addressing each copy appropriately, then forwarding each copy to its final destination. Although this solution requires a greater communications channel bandwidth it is frequently tolerated because determination of an optimal solution yielding a minimal number of transmissions is NP-complete. This paper proposes a resolution to the problem by employing a Hopfield neural network. Results are compared against optimal solutions derived through exhaustive search.