Gradual neural network approach for broadcast scheduling in packet radio networks

A gradual neural network (GNN) approach with the improved gradual expansion scheme is proposed for broadcast scheduling in packet radio (PR) networks. A PR network provides data communications services to geographically distributed nodes through a radio channel. A time division multiple access (TDMA) protocol is adopted for the network. Packets are transmitted in repetition of a TDMA cycle, where the delay time in packet broadcasting should be minimized. The proposed gradual expansion scheme resolves the constraints of the problem using neuron inputs and outputs to reduce the computation time. Besides, an additional slot to a TDMA cycle is considered for slot assignments when a valid solution is not obtained within a current cycle. The performance comparison with a conventional GNN and a greedy algorithm shows the effectiveness of the proposed GNN approach.