As　urbanization　progresses,　train-to-train　(T2T)　communication　for　communication-based　train　control　(CBTC)　systems　has　become　crucial,　leading　to　mobile　ad　hoc　networks　(MANETs)　between　trains.　The　dynamic　network　topology　poses　challenges　to　the　high　reliability　needed　for　safe　rail　transportation　information.　To　address　this,　we　propose　a　cooperative　medium　access　control　(MAC)　communication　scheme　based　on　distributed　time　division　multiple　access　(DTDMA)　for　T2T　communications.　This　scheme　uses　multiple　nodes　in　their　idle　time　slots　to　cooperatively　forward　failed　data.　To　ensure　the　reliability　and　effectiveness　of　this　cooperation,　we　propose　a　cooperative　MAC　protocol　and　frame　structures　combined　with　the　automatic　repeat　request　(ARQ)　mechanism　for　reliability　and　effectiveness.　Additionally,　we　propose　a　cooperative　node　selection　method,　modeled　as　a　Markov　decision　process　(MDP),　to　maximize　system　throughput　and　minimize　overhead.　For　actions　with　hybrid　space,　we　use　the　soft　actor-critic　(SAC)　to　set　a　continuous　threshold　and　select　nodes　discretely.　Simulations　show　that　our　proposed　cooperative　MAC　protocol　and　cooperative　node　selection　method　effectively　enhance　transmission　reliability.　　©　1967-2012　IEEE.