A Reinforcement Learning Approach for Efficient Opportunistic Vehicle-to-Cloud Data Transfer

Abstract: Vehicular crowdsensing is anticipated to become a key catalyst for data-driven optimization in the Intelligent Transportation System (ITS) domain. Yet, the expected growth in massive Machine-type Communication (mMTC) caused by vehicle-to-cloud transmissions will confront the cellular network infrastructure with great capacity-related challenges. A cognitive way for achieving relief without introducing additional physical infrastructure is the application of opportunistic data transfer for delay-tolerant applications. Hereby, the clients schedule their data transmissions in a channel-aware manner in order to avoid retransmissions and interference with other cell users. In this paper, we introduce a novel approach for this type of resourceaware data transfer which brings together supervised learning for network quality prediction with reinforcement learningbased decision making. The performance evaluation is carried out using data-driven network simulation and real world experiments in the public cellular networks of multiple Mobile Network Operators (MNOs) in different scenarios. The proposed transmission scheme significantly outperforms state-of-the-art probabilistic approaches in most scenarios and achieves data rate improvements of up to 181% in uplink and up to 270% in downlink transmission direction in comparison to conventional periodic data transfer.