Vehicular cloud computing is a new paradigm that integrates VANET and cloud computing. It refers to the vehicles' use of outside resources: they access to the conventional cloud or cooperate with other vehicles by sharing their resources (i.e. formation of a temporary vehicular cloud). Both these two ways induce different costs such as expensive cellular links in the case of access to the conventional cloud and intermittent links in the case of vehicles' cooperation. In this paper, we focus on the proactive decision made by the vehicle (i.e. choose the access to the vehicular cloud or to the cloud computing), where this decision depends on many parameters hardly to control. In other words, our goal is to determine the best technology to access the cloud (i.e wifi/LTE). We propose to model the evolution of the vehicles' strategy as an evolutionary game. To avoid the utilization of a centralized controller, we propose a distributed Q-Learning based Vehicular Cloud Access algorithm (QL-VCA) that allows each vehicle to select the way of access independently. The simulation results show that the proposed Q-learning algorithm provides users with better throughput and lower packet loss rate compared to the other scenarios (i.e. using only the LTE or the wifi).