To　solve　the　problem　of　redundant　exploration　in　the　recursive　expansion　of　the　Fast　Marching　Tree　algorithm　(FMT*),　a　Direction　Selection-based　FMT*　algorithm　(DS-FMT*)　was　proposed.　The　algorithm　generated　a　uniform　distribution　of　direction　selection　lines　around　the　extended　sample　to　judge　the　surrounding　obstacles　and　select　the　direction　favorable　for　expansion　as　the　candidate　exploration　direction.　Then,　the　actual　exploration　direction　of　the　sample　to　be　expanded　to　the　next　sample　was　compared　with　the　candidate　exploration　direction.　If　the　angle　between　the　actual　exploration　direction　and　the　candidate　exploration　direction　was　consistent,　the　sample　in　this　direction　would　be　given　priority.　Meanwhile,　combined　with　the　cost　comparison　of　samples,　the　recursive　expansion　process　of　FMT*　was　improved　to　reduce　the　computational　complexity.　The　proposed　DS-FMT*　algorithm　was　compared　with　other　advanced　similar　algorithms,　which　proved　that　DS-FMT*　could　improve　the　planning　efficiency　while　ensuring　the　good-quality　of　the　path.　Besides,　the　effectiveness　of　the　proposed　algorithm　was　verified　by　practical　application　experiments　of　mobile　robot　path　planning.　©　2021,　Editorial　Department　of　CIMS.　All　right　reserved.