Robot　navigation　under　dynamic　environment　is　considered　as　one　of　the　key　indications　of　machine＇s　intelligence.　This　paper　proposed　a　program　learning　algorithm　for　robot　navigation,　so　as　to　efficiently　increase　the　intelligence　of　the　machine.　The　Tolman　mouse　maze　experiment　was　expanded　to　help　the　robot　to　develop　intuitive　intelligence,　that　it　can　lean　the　regularity　the　environment　change.　Firstly,　the　efficient　environment　map　is　constructed　by　SOM　based　on　the　cloud　point　generated　randomly　according　to　the　environment.　Then　the　robot　will　automatically　choose　how　to　reach　the　goal　destination　according　to　the　shortest　path　generated　by　A　star　algorithm　formerly.　And　corresponding　reactions　of　the　robot　are　determined　by　cycling　time,　which　is　the　program　we　help　the　robot　realize　which　path　it　is　supposed　to　select.　The　validity　of　the　proposed　algorithm　is　proved　by　MATLAB　and　gazebo　of　ROS　indigo,　besides,　we　also　did　some　physical　experiments　which　have　well　shown　the　robots＇　highly　developed　intelligence　of　finding　the　shortest　path　rapidly.　©　2019　IEEE.