Blind collision resolution for wireless multiple access using independent component analysis

Abstract

In wireless random access networks, packet collision is an important medium access problem for the packet radio environment. Simple protocols such as the slotted ALOHA do not provide an efficient solution to the collision problem since they generally suffer from a severe throughput penalty and the underutilization of the channel resources. Thi~ss,i gnal separation techniques used in signal processing can be employed to improve the throughput and delay performance of the system. Network-assisted diversity multiple access (NDMA) and blind network-assisted diversity multiple access (BNDMA) are two such techniques. User specific ID sequences are necessary to be inserted to the transmitted data packets to resolve the collision by NDMA. BNDMA does not need this requirement, and collision is resolved by using parallel factor (PARAFAC) analysis. In this thesis, independent component analysis (ICA) is integrated into the multiple access mechanism to provide a significant improvement in terms of system throughput and the transmission delay. Performance comparisons with NDMA and BNDMA are furnished. The number of unresolved packet retransmissions is controlled by the base station and can be decreased through spatial diversity.