DLB-SDPOP: A Multiagent Pseudo-tree Repair Algorithm for Load Balancing in WLANs

The traffic load of wireless local area networks (WLANs) is often distributed unevenly among access points. In addition, interference from collocated wireless devices operating in the same unlicensed frequency band may cause WLANs to become unstable, leading to temporary failures of access points. This paper addresses the questions of how to dynamically balance the load and how to quickly respond to instability in WLANs. We present a new decentralized multi-agent load balancing framework for WLANs that uses DLB-SDPOP, a distributed optimization algorithm, to dynamically load balance the WLAN. The algorithm implements a less expensive pseudo-tree repair mechanism instead of complete pseudo-tree reconstruction when instability problems occur. It also leverages an efficient communication mechanism to control communication overhead in heavily loaded situations. We empirically show that DLB-SDPOP improves WLAN load balancing performance significantly.