With　the　development　of　intelligent　manufacturing　technology,　automatic　guided　vehicle　(AGV)　has　been　widely　used　in　workshop　logistics　transportation.　However,　the　complex　and　changeable　production　process　requires　efficient　and　dynamic　transportation　and　distribution.　In　this　paper,　based　on　the　results　of　process　scheduling,　a　method　of　independent　decision-making　transportation　process　is　studied.　A　transportation　strategy　training　method　with　breakpoint　continuation　and　hierarchical　feedback　is　proposed　based　on　deep　Q-network　(DQN).　Transportation　scheduling　can　be　quickly　decided　and　adjusted　to　adapt　to　dynamic　and　changeable　orders.　The　method　is　applied　to　the　FJSP　problem,　and　the　data　experiments　show　the　effectiveness　of　the　method.　©　2021　IEEE.