Multimodal dynamic multiple access (MDMA) in wireless packet networks

Abstract

We investigate novel channel access schemes for packetized wireless networks, which can dynamically switch between distinct transmission modes in order to better match the channel state and deliver packets to the receiver with higher success probability (rate). We call them multimodal dynamic multiple access-MDMA schemes. Based on the observed channel impairment state (typically a combination of interference, fading, multipath, etc.) and the transmitter queue packet backlog at any time instant (slot), each user autonomously selects the best transmission mode to activate and power level to transmit at. First, a general formulation of the MDMA problem is introduced in several methodological steps of progressive complexity. It is based on dynamic programming and captures the basic tradeoffs. Analytical issues are not pursued in detail here, but instead several ubiquitous structural properties of MDMA schemes are identified and explored. Based on those, a novel suite of MDMA algorithms is designed and evaluated. On a simulated baseline scenario, MDMA is shown to achieve over 30% higher throughput than previously studied raw PCMA schemes and even higher performance gains over other standard benchmark ones. This indicates that MDMA schemes can release 'latent' network capacity which is suppressed by others, and should be further explored. This study is a first step towards designing full MDMA protocols for high-performance wireless packet networks.