Obstacle　avoidance　behavior,　as　an　important　part　of　pedestrian　evacuation,　is　of　great　significance　to　the　study　of　evacuation　efficiency.　This　paper　investigates　the　definitions　of　desired　direction　in　the　social　force　model　for　obstacle　avoidance　of　two　types　of　pedestrians,　including　pedestrians　with　and　without　complete　evacuation　information.　Ant　colony　optimization　algorithm　is　adopted　to　navigate　pedestrians　with　complete　information.　Herding　behavior,　individual　preference　affected　by　obstacles　and　walls　are　taken　into　consideration　when　defining　the　desired　direction　of　pedestrians　with　local　information.　Simulation　experiments　are　carried　out　to　explore　obstacle　avoidance　dynamics,　the　effects　of　herding　behavior　and　visibility.　Results　indicate　that　pedestrians　with　complete　information　can　not　only　avoid　obstacles　better,　but　also　have　the　ability　to　choose　the　shorter　route　to　the　exit　which　could　improve　the　leaving　efficiency.　Meanwhile,　the　trajectories　of　pedestrians　with　local　information　are　always　accompanied　by　some　twists　and　turns,　which　could　obviously　lead　to　a　waste　of　time.　No　matter　whether　the　proportion　of　pedestrians　with　complete　information　is　large　or　not,　herding　behavior　can　make　the　trajectories　of　pedestrians　with　local　information　smoother,　and　the　individual　behavior　can　make　their　trajectories　more　curved.　Moreover,　the　larger　the　visibility　radius　is,　the　smoother　trajectories　become　and　the　greater　effective　displacements　for　pedestrians　with　local　information　are.　(C)　2021　Published　by　Elsevier　B.V.