Computation offloading strategy based on deep reinforcement learning for connected and autonomous vehicle in vehicular edge computing

Abstract Connected and Automated Vehicle (CAV) is a transformative technology that has great potential to improve urban traffic driving safety. Electric (EV) becoming the key subject of next-generation CAVs by virtue its advantages in energy saving. Due limited endurance computing capacity EVs, it challenging meet surging demand for computing-intensive delay-sensitive in-vehicle intelligent applications. Therefore, computation offloading been employed extend single vehicle’s capacity. Although various strategies have proposed achieve good performace Vehicular Edge Computing (VEC) environment, remains jointly optimize failure rate total consumption process. To address this challenge, paper, we establish model based on Markov Decision Process (MDP), taking into consideration task dependencies, vehicle mobility, different resources offloading. We then design strategy deep reinforcement learning, leverage Deep Q-Network Simulated Annealing (SA-DQN) algorithm joint objectives. Experimental results show effectively reduces application