Multi-user MIMO and Resource Allocation for Multi-Access Point Wi-Fi Systems

The increasing demands of emerging applications such as virtual and augmented
reality, online gaming, and industrial wireless services have accelerated the evolution
of next-generation Wi-Fi standards toward higher throughput, lower latency, and
improved reliability. IEEE 802.11be (Wi-Fi 7) introduces Extremely High Throughput
(EHT), while IEEE 802.11bn is being developed to support Ultra-High Reliability
(UHR), where multi-access point (multi-AP) coordination has emerged as a key
enabling technology. Among the four coordination schemes, this thesis focuses
primarily on coordinated orthogonal frequency division multiple access (C-OFDMA)
under practical Wi-Fi system assumptions.
To establish the research foundation, the CSMA/CA channel access mechanism
is reconstructed and a fairness-based multi-agent reinforcement learning framework
is reproduced to examine its ability to reduce channel access randomness and latency
in dense environments. Building on this, a fixed resource unit (RU) allocation
strategy is proposed for homogeneous coordinated Wi-Fi systems, together with a
joint C-OFDMA and coordinated spatial reuse (C-SR) scheme for interference-aware
throughput improvement. The study is then extended to heterogeneous systems
where stations experience unequal channel gains. For this case, a deep reinforcement
learning (DRL)-based variable RU allocation algorithm is developed under max-min
fairness, demonstrating clear improvement in minimum throughput and Jain’s
fairness compared with benchmark allocation methods. Under the same system
model, proportional fairness is investigated and a closed-form RU allocation solution
is derived.
Finally, a realistic online scheduling problem is considered where both packet
arrival rates and channel gains vary over time. To address the NP-hard queue-aware
RU allocation problem, a DRL-based scheduling algorithm is proposed to minimise
queueing delay, reduce packet drops, and shorten transmission opportunity duration.
Simulation results confirm that the proposed methods improve fairness, reliability,
and delay performance, while maintaining practical suitability for next-generation
coordinated Wi-Fi systems.