Imola: A decentralised learning-driven protocol for multi-hop White-Fi

In this paper we tackle the digital exclusion problem in developing and remote locations by proposing Imola, an inexpensive learning-driven access mechanism for multi-hop wireless networks that operate across TV white-spaces (TVWS). Stations running Imola only rely on passively acquired neighbourhood information to achieve scheduled-like operation in a decentralised way, without explicit synchronisation. Our design overcomes pathological circumstances such as hidden and exposed terminals that arise due to carrier sensing and are exceptionally problematic in low frequency bands. We present a prototype implementation of our proposal and conduct experiments in a real test bed, which confirms the practical feasibility of deploying our solution in mesh networks that build upon the IEEE 802.11af standard. Finally, the extensive system level simulations we perform demonstrate that Imola achieves up to 4× more throughput than the channel access protocol defined by the standard and reduces frame loss rate by up to 100%.