PROJECT TITLE :

A Novel Deep Reinforcement Learning based Relay Selection for Broadcasting in Vehicular Ad hoc Networks

ABSTRACT:

It is believed that VANETs, also known as Vehicular Ad hoc NETworks, are among the largest networks in the world. These networks are offering a wide variety of services, including video on demand, driver assistance, safety services, infotainment applications, and safety services. On the one hand, VANETs are distinguished by their ad hoc topology and dynamic behavior, the former of which varies depending on the urban setting, and the latter of which is subject to significant shifts on highways. On the other hand, disseminating information is an essential activity that must be completed in order to provide multiple services. In light of this, the task of broadcasting is a difficult problem that requires further investigation. In point of fact, artificial intelligence and computing that is based on learning appear to be one of the most appropriate options that best fits the dynamic behavior of VANETs. This is the task that needs to be completed. In light of this, the purpose of this paper is to propose a novel technique for hybrid relay selection that is based on a method of reinforcement learning and is used to perform the broadcasting task. In the first phase of our proposal, we will combine a classification that is based on an artificial neural network and will be applied to certain forwarding nodes. In the second phase, we will apply the Viterbi algorithm as a reinforcement tool in order to refine the initial classification. We use a grid map scenario that has varying traffic densities as the basis for our evaluation of the effectiveness of our contribution. After that, we conduct an analysis and make a comparison of the results of the simulation with those of other methods found in the existing body of research based on various parameters such as the percentage of successful transmissions, the amount of data that is lost, the number of rebroadcasts that are saved, and the amount of delay. In conclusion, we demonstrate that the proposed method that combines Deep Learning with reinforcement learning outperforms other recently proposed broadcasting schemes based on the results, which show that the new solution increased the success rate by 16%, the saved rebroadcasts by 20%, and reduced the delay by 23%. These improvements were made possible by combining deep learning with reinforcement learning.