A reinforcement learning approach to optimize downloads over mobile networks

Dedicated Short Range Communication is attracting a lot of interest these days due to its utility in vehicular safety applications, intelligent transportation system and infotainment applications. Such vehicular networks are characterized by the highly dynamic changes in topology, no significant power constraints and ephemeral links. Considering an interaction between the client and server nodes that last for a random duration of time, an important question is to maximize the amount of useful content downloaded by the client, either in a single request phase, or iteratively in multiple phases. The aim of this work is to propose and investigate a multiphase request model using Markov Decision Process and compare its efficiency against a single phase version. We show that a multiphase request protocol performs better than single phase protocol.