This　paper　presents　a　new　method　for　lunar　rover　path　planning　and　obstacle　avoidance　in　virtual　triaxiality　terrain.　The　method　takes　into　account　the　two-dimension　configuration　space　and　gradient　information.　Firstly,　through　analyzing　the　obstacle　and　terrain＇s　geometry　characteristic,　obtain　the　obstacle　classification　map　and　mean　slope　map　according　with　the　capacity　of　rover＇s　climbing　and　the　motion　bound.　It　is　the　obstacle　whose　grade　is　larger　than　the　rover＇s　climbing　capacity　and　it　is　the　free　district　whose　grade　and　area　are　both　less　than　the　rover＇s　guideline.　Secondly,　superimpose　a　2D　Cartesian　grid　on　the　dynamic　window.　The　value　of　grid　is　directly　proportional　to　the　location＇s　gradient,　directly　proportional　to　the　area　of　obstacle　occupied,　and　inversely　proportional　to　the　distance　from　the　grid　to　the　goal.　The　value　would　be　as　the　input　of　path　planning.　Lastly,　plan　the　path　using　the　behavior　control　method　in　which　the　motion　direction　is　towards　the　grid　whose　value　is　minimum.　By　the　behavior　selector,　determine　the　appropriate　behavior　for　the　rover　to　perform　given　its　current　state　and　information　from　local　and　global　cost　functions.　There　are　two　basic　behaviors　which　are　the　moving　to　goal　behavior　and　obstacle　avoidance　behavior.　Thereinto　the　moving　to　goal　behavior　adopts　fuzzy　control　method　and　the　obstacle　avoidance　behavior　uses　VFH　method.　Additionally,　add　the　dynamics　and　kinematics　restrictions　to　control.　The　method　has　been　tested　in　triaxiality　environment　simulation　and　the　results　support　the　validity　of　the　method.