Exit　selection　behavior　is　important　for　evacuation　processes　and　safe　facility　design.　Pedestrians＇　ability　to　select　an　exit　route　is　affected　by　many　factors,　and　will　also　impact　on　evacuation　efficiency.　It　should　not　be　neglected　when　modeling　evacuation.　In　the　case　of　asymmetric　exits　and　distribution　of　pedestrians,　an　important　issue　is　how　to　identify　a　pedestrian＇s　target　when　performing　simulations.　However,　this　has　not　been　well　investigated,　especially　in　discrete　models.　In　this　regard,　we　tend　to　investigate　pedestrian　exit　choice　behavior　by　integrating　the　least　effort　algorithm　with　a　cellular　automaton　model.　The　distance　to　exits　and　crowd　density　around　exits　are　involved.　Simulations　are　conducted　in　a　two-exit　room.　Evacuation　time　in　scenarios　where　there　are　some　special　distributions　of　pedestrians　is　compared　with　that　in　scenarios　where　pedestrians　are　randomly　distributed.　The　influence　of　the　weighted　value　of　the　distance　to　exits　or　crowd　density　around　exits　on　evacuation　time　is　also　studied.　Useful　suggestions　are　provided.　The　effect　of　locations　of　exits　on　evacuation　time　and　the　cumulative　number　of　egress　pedestrians　are　further　investigated.　Results　demonstrate　that　two　exits　located　in　different　walls,　especially　in　symmetrical　walls,　are　helpful　in　evacuation　processes.　This　result　is　in　line　with　that　in　other　references.　It　is　hoped　that　this　work　will　be　helpful　in　improving　evacuation　rules　of　discrete　models　in　multi-exit　situations.　©　2017　The　Author(s).