A Hybrid Approach for Optimal Cluster Head Selection in VANETs with Various Topologies Using Fuzzy Logic, Moth Flame Optimization, and Machine Learning

Nowadays, the Vehicular Ad hoc Network (VANET) technology is used to improve the quality of the transportation systems and road safety between Vehicles (Vs). The main routing challenges in VANETs include their dynamic and unstable structure, energy limitations of the Vs, and the use of intermediate Vs. Clustering is used to balance the overhead, increase lifetime, and enhance data collection in VANETs. Finding the optimal Cluster Head (CH) is an NP-hard problem, and heuristic and metaheuristic methods are often employed to solve it. In this paper, we propose a method for routing and optimal CH selection among all. In each cycle and across all VANETs with different topologies, various Vs features are first collected, and then using a heuristic method, the Fuzzy Inference System (FIS), the Optimization Fitness Function ( OFF ) value of all Vs is calculated to determine the optimal CHs. Additionally, the metaheuristic Moth Flame Optimization (MFO) algorithm is used to tune and set the coefficients and rules of the FIS. Finally, to train and test VANET behavior patterns across various topologies, Decision Trees (DTs) based on the Random Forest (RF) ensemble Machine Learning (ML) method are utilized. Simulation results show that the proposed method outperforms clustering-based routing protocols such as LEACH, AODV, DSRC, CBRP, and GPSR in VANETs in terms of the number of alive and dead Vs, average network lifetime, routing overhead, end-to-end delay, throughput, packet delivery rate, and execution time.