In　this　paper,　we　propose　a　scan-based　path　planning　algorithm,　in　which　scanning　method　is　introduced　to　search　obstacle＇s　tangent　and　structure　topology　map　on-line,　independent　of　the　geometric　shapes　and　vertexes＇　information　of　the　obstacles.　According　to　the　structured　partial　topology　map　of　the　environment　and　the　global　information,　heuristic　function　is　used　to　select　tangent　point　to　expand,　while　update　and　complement　the　structured　topology　map,　till　a　traversable　path　between　starting　point　and　goal　is　found　in　the　topology　map.　Two　behavioral　modes,　＇go　straight　line＇　and　＇boundary　following＇,　are　included,　and　switch　between　them　guarantees　that　the　robot　can　achieve　the　goal　safely.　While　a　direction　selection　method　is　used　to　minimize　scanning　scope　and　improve　efficiency.　The　simulation　results　prove　the　effectiveness　and　correctness　of　this　algorithm.　In　short,　this　algorithm　has　lower　computational　complexity　compared　with　other　existed　roadmap　algorithms,　is　feasible　for　obstacle　with　arbitrary　shape,　and　easy　to　realize　on-line.　©　2005　IEEE.