Machine Learning based Routing approach and Resource Management in Vehicular Adhoc Networks

Vehicular Adhoc Network (VANET) is one of the innovative research areas in Adhoc network that establishes communication between participating nodes with help of inter-mediate node or external access point. Since mobility of vehicles in this network is very high, structure and communiqué links are habitually modified. Identifying way of reducing latency issue and improve system throughput are two major issues. Great majority of task depends on motion planning and choosing ways with increased vehicle velocity for transmission of packet, eliminating carry-and-forward scenarios and reducing system throughput. However, even though vehicle densities change so quickly, using such methods, won't get reliable results over the period of time transportation. Estimate necessary information for routing protocols, we suggest new route request scheme, which we call Machine Learning-Contributed Path Selection (MLCPS) system in this paper. Machine learning is used to maintain road information in roadside units (RSU), predict vehicle movements and then select appropriate routing paths with relatively high transmitting capability for packet forwarding. Furthermore, MLCPS can aid in trying to determine the transmitting path between two RSUs based on the expected destination position and both directions' approximate transmission delays. Result shows significant improvement in parameters like packet delivery ratio and end to end delay.