This paper presents a self-organizing time division multiple access (SO-TDMA) protocol for contention resolution aiming to support delay-sensitive applications. The proposed SO-TDMA follows a cognition cycle where each node independently observes the operation environment, learns about the network traffic load, and then makes decisions to adapt the protocol for smart coexistence. Channel access operation in SO-TDMA is similar to carrier-sense multiple-access (CSMA) in the beginning, but then quickly converges to TDMA with an adaptive pseudo-frame structure. This approach has the benefits of TDMA in a high-load traffic condition, and overcomes its disadvantages in low-load, heterogeneous traffic scenarios. Furthermore, it supports distributed and asynchronous channel-access operation. These are achieved by adapting the transmission-opportunity duration to the common idle/busy channel state information acquired by each node, without any explicit message passing among nodes. The process of adjusting the transmission duration is modeled as a congestion control problem to develop an additive-increase-multiplicative-decrease (AIMD) algorithm, which monotonically converges to fairness. Furthermore, the initial access phase of SO-TDMA is modeled as a Markov chain with one absorbing state and its required convergence time is studied accordingly. Performance of SO-TDMA in terms of effective capacity, system throughput, collision probability, delay-outage probability and fairness is investigated. Simulation results illustrate its effectiveness in performance improvement, approaching the ideal case that needs complete and precise information about the queue length and the channel conditions of all nodes.
Read full abstract