The vehicular ad-hoc networks (VANETs) support wireless communication among moving vehicles, infrastructures as well as other devices. In VANETs, the problem of sharing the same channel is complex, which results in more packet collisions in resource allocation unless the resource information is unified for each vehicle. The process of resource allocation among vehicles must be optimized for efficiently using the possible wireless bandwidths and the successful configuration of VANETs. For efficient resource allocation, we apply Q-learning that allows many vehicles in a network, which can make the process of exchanging data among them more efficient. The policy of choosing contention window size can be learned, where a hybrid linear and exponential contention window size adjustment is considered. Vehicles learn in the process of maximizing successful transmission of data packets and minimizing bandwidth waste. Furthermore, the proposed algorithm performs better than existing back-off algorithms.
This study was conducted as a result of the research of the S W-oriented university project of the Ministry of Science and ICT and the Ministry of Information and Communication lP anning and Evaluation (2015-0-0008) and the National Res earch Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (No. 201R1A2C10085 0).