Machine Learning approach for task offloading strategy in IoV.

As vehicles continue to integrate with the Internet of Things to access better services, the abundance of vehicular applications and network fluctuations pose a challenge to in-vehicle terminals to perform efficient computation. To tackle this issue, a task offloading algorithm for the Internet of Vehicles (IoV) was developed using a four-stage local edge cloud and reinforcement learning model. The algorithm involved implementing communication methods and cost functions between vehicles, analyzing their computational requirements and real-time status-based cost function, and proposing an experience-based offloading strategy using multi-agent reinforcement learning. The simulation results demonstrated that the algorithm improved the likelihood of task success while balancing task vehicle utility, and service vehicle utility under different constraints.