Orthogonal Resource Allocation Using SVM for CSMA/CA

Abstract

Due to the wide spread of machine-to-machine (M2M) communications and Internet-of-things (IoT), large number of wireless terminals are densely deployed. Under such dense deployment, it is necessary to manage the mutual interference among wireless terminals. Carrier-sense multiple access/collision avoidance (CSMA/CA) is one of the random access schemes that allow wireless terminals to access a channel while avoiding such mutual interference. However, if two wireless terminals are in the hidden terminal relation, a packet collision may happen as the carrier sense mechanism does not work. This results in the degradation of packet delivery rate (PDR) performance. Whether or not particular two wireless terminals are in the hidden terminal relation is an unobservable information from a network controller such as an access point (AP). In this paper, we tackle this problem by using machine learning. By using machine learning, the wireless controller makes a guess of the unobservable information from the observable information. We will use the wireless terminal locations and received signal strength at APs as the observable information. Based on the estimated unobservable information, the wireless controller assigns orthogonal resources to the wireless terminals that are in the relationship of hidden terminal in order to avoid the packet collision. Numerical results confirm that the proposed approach can improve the PDR performance up to 15% compared to the random resource allocation scheme.