Multi-AP Multi-Dimensional Resource Coordination for Next-Generation Wi-Fi Networks via Multi-Agent Deep Reinforcement Learning

Abstract

The ever-growing Wi-Fi data traffic and diverse quality-of-service (QoS) requirements of coexisting stations (STAs) exacerbate the challenges of coordinated resource sharing and spatial-temporal interference mitigation among neighboring access points (APs). This paper proposes a novel multi-AP multi-dimensional resource coordination mechanism to intelligently match the distributed resources with the STAs’ heterogeneous QoS demands under a standard-compliant protocol. Specifically, the proposed method allows collocated APs to dynamically orchestrate multi-dimensional resources by jointly enabling resource unit (RU)-level spatial reuse and spectral-temporal resource sharing through coordinated orthogonal frequency-division multiple access (Co-OFDMA). To achieve the coordination objective of maximizing collective network utility, a hierarchically structured two-stage solution is designed to tackle the NP-hardness and non-stationarity of the decision variables. Stage-I lets APs evaluate RU reusability for adaptive configuration of channel access sensitivity through finer-grained interference assessment. During Stage-II, APs leverage the agent-based learning capabilities to collaboratively exploit the varying network-wide interference patterns and QoS characteristics. Particularly, a multi-agent hybrid QMIX algorithm with parameterized actor networks (MA-HQPAN) is proposed to solve the RU assignment and transmit power allocation problem. Simulations verify that the proposed method can opportunistically coordinate multi-dimensional resources for diverse QoS provisioning, outperforming baselines by robust and effective interference mitigation with fairness guarantees.