With　the　development　of　smart　vehicles,　more　and　more　multimedia　contents　are　generated　near　the　vehicles.　Edge　caching　becomes　an　inevitable　method　of　caching　content　near　the　vehicles　to　reduce　the　overhead　of　core　network　and　improve　the　quality　of　service.　However,　due　to　the　movement　of　vehicles,　the　contents＇　popularity　may　vary　over　time.　Complex　content　delivery　and　high　mobility　of　vehicles　present　new　challenges　to　edge　caching　in　dynamic　environments.　In　this　paper,　a　cooperative　caching　scheme　which　exploit　Hawkes　process　to　predict　content　popularity　is　proposed.,　which　considers　the　freshness　and　timeliness　of　content　requests.　A　deep　reinforcement　learning　(DRL)　is　used　to　maximize　the　caching　efficiency　which　considers　system　throughput　and　caching　energy　consumption.　Simulation　results　show　that　the　proposed　strategy　can　get　higher　benefit　than　other　strategies.　©　2021　IEEE.