An Online Evolutionary Programming Method for Parameters of Wireless Networks

Abstract

Wireless networks operate in rapidly changing environments. Often parameters for particular algorithms are set with particular environments in mind, or assume certain conditions. When conditions change from interference, user mobility, handover and changing demand, the network may be unable to cope. To solve some of these problems we propose an online evolutionary approach to parameter computation. The online approach allows for quick computation of new parameter values while still retaining some history of past actions. We apply this approach to mixed bias scheduling and demonstrate that the approach works well when compared with existing mixed bias approaches and IEEE 802.11 DCF. The evolutionary programming approach achieves a significantly reduced end-to-end delay while maintaining comparable packet delivery ratio when evaluated using simulation experiments.