Root-MUSIC-based methods for blind network-assisted diversity multiple access
Access full-text files
Date
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Packet collisions in wireless networks degrade the throughput and impede the system performance. The collided packets are typically corrupted and get discarded. Channelization methods avoid collisions through fixed assignment of communication resources to the system users, but they do not take into account the randomness of packet arrivals. Statistical multiplexing optimally adapts the allocation of resources to the instantaneous traffic demands of the users. However, it is only possible in the downlink wherein the data streams are managed by one station. Random-access methods mimic statistical multiplexing by dynamically assigning resources to users. A slot is wasted if the channel incurs a collision, and the collided packets have to be retransmitted.
First, we present a cross-layer design for providing multiple access to a shared wireless link. While retransmissions are controlled by the medium access control (MAC) layer, this creates sufficient diversity to recover the collided packets in the physical (PHY) layer. Both the number and identities of the involved transmitters in a collision are unknown to the receiver. The signal separation is done blindly using root-MUSIC-like algorithms. We solve the collision resolution problem in four network-operation modes: synchronous blocking mode, synchronous non-blocking mode, asynchronous blocking mode and asynchronous non-blocking mode.
Second, we evaluate the decoding performance of the algorithms in block-fading channels with additive white Gaussian noise. We analytically demonstrate the effect of signal-to-noise ratio and the number of retransmissions on the signal separation capability of the proposed methods for a given number of collided packets.
Third, we evaluate the network throughput and mean packet queueing delay for the proposed collision resolution algorithms analytically and numerically. We derive conditions for stability of the queueing network as function of the mean packet arrival rates.