Fair Randomized Antenna Allocation in Asynchronous MIMO Multi-Hop Networks

Abstract

Previous work has shown that CSMA-based protocols such as the IEEE 802.11 can yield flow starvation in multi-hop wireless networks. While prior protocol designs alleviated such starvation by utilizing MIMO, they require global synchronization to temporally align transmissions and require per-packet distribution of the complete channel state information. In this paper, we experimentally show that MIMO networks based on pre-802.11 n medium access, the state-of-the-art for asynchronous MIMO CSMA, worsen starvation as compared to SISO networks. Consequently, we design an asynchronous MIMO MAC protocol that counters starvation. We show that randomized and non-greedy antenna allocation coupled with local residual capacity estimation results in previously-starving nodes capturing a fair share of system resources while simultaneously exploiting throughput gains available to multi-antenna systems.