Optimal broadcast scheduling in packet radio networks using mean field annealing

We present an efficient broadcast scheduling algorithm based on mean field annealing (MFA) neural networks. Packet radio (PR) is a technology that applies the packet switching technique to the broadcast radio environment. In a PR network, a single high-speed wideband channel is shared by all PR stations. When a time-division multi-access protocol is used, the access to the channel by the stations' transmissions must be properly scheduled in both the time and space domains in order to avoid collisions or interferences. It is proven that such a scheduling problem is NP-complete. Therefore, an efficient polynomial algorithm rarely exists, and a mean field annealing-based algorithm is proposed to schedule the stations' transmissions in a frame consisting of certain number of time slots. Numerical examples and comparisons with some existing scheduling algorithms have shown that the proposed scheme can find near-optimal solutions with reasonable computational complexity. Both time delay and channel utilization are calculated based on the found schedules.