With the advocacy of green renewable energy, Electric Vehicles (EVs) have gradually become the mainstream in the automobile market. Due to the finite edge resources of the Internet of EVs, this paper integrates idle communication, caching and computational resources of EVs to enrich the available resources for vehicular task migration. Considering the limited capacity and resources of EVs, a distributed lightweight imitation learning-based efficient Task cOoperative migration Policy Integrating 3C resource policy, named TOPIC, is proposed to maximize the obtained quality of service. The experimental results based on the real-world traffic dataset of Hangzhou (China) demonstrate the QoS obtained based on the expert policy and agent policy of TOPIC is about 3 times higher than other representative policies.