Mobility-Aware Adaptive Channel Contention for Delay-Sensitive Communications

Distributed channel access mechanisms such as carrier sense multiple access (CSMA), adopted in Wi-Fi networks, lack centralized coordination and global queue state information. Thus, they are inefficient in supporting delay-sensitive mobile traffic. For this problem, we introduce an adaptive channel contention mechanism aimed at improving queuing performance in a distributed millimeter-wave (mmWave) uplink system equipped with environment and mobility sensing capabilities. Specifically, mobile agents adjust their back-off timer parameters based on locally available information, including uplink queue lengths, current channel quality, and predicted channel statistics, where predictions are enabled by environmental and mobility sensing. We formulate the queuing optimization problem under this adaptive mechanism as a decentralized multi-agent Markov decision process (Dec-MA-MDP). Although contention decisions are made locally by each mobile agent, the corresponding policies for all agents are optimized centrally using system-wide statistics obtained prior to scheduling. In the proposed method, local policies are approximated via analytical models, reducing the policy optimization to a stochastic optimization problem defined along an adaptive Markov chain. To enhance efficiency, we further propose an unbiased gradient estimator, enabling policy refinement through stochastic gradient descent. Simulation results for both indoor and outdoor scenarios show that the proposed method achieves significantly higher optimization efficiency than conventional methods, such as simultaneous perturbation stochastic approximation (SPSA).