Since　the　increasing　requirement　of　indoor　navigation　with　the　assurance　of　safety,　Deep　Learning　was　been　used　in　this　area　by　numerous　scholars.　In　order　to　help　unmanned　aerial　vehicle　navigate　in　various　environment　without　prior　experience,　this　paper　proposed　the　method　of　Deep　Q　Learning　with　the　Squeeze-and-Excitation　Networks　and　Dense　blocks.　The　structure　of　the　Q-Network　also　amended　with　the　information　of　LIDAR　processed　by　Long　Short-Term　Memory　to　make　sure　the　accuracy.　The　structure　of　the　network　shows　in　figure　6.　The　experiment　conducted　in　simulated　environment.　Experimental　results　indicated　that　this　method　excelled　in　efficiency　and　accuracy　than　other　classic　Deep　Q　Networks　like　classic　Deep　Q　Network　and　Double　Deep　Q　Network.　The　results　displaying　in　figure　8　and　figure　9　shows　unmanned　aerial　vehicle　can　arrive　at　target　place　more　accurately　than　other　networks.　©　Published　under　licence　by　IOP　Publishing　Ltd.