This　paper　presents　an　efficient　Simultaneous　Localization　and　Map-Building　(SLAM)　technique　for　Lunar　Rover　navigation　based　on　laser　scan-matching　and　Rao-Blackwellized　Particle　Filter　(RBPF).　It　is　an　improved　version　of　FastSLAM　algorithm　which　deals　with　unstructured　environment.　As　inconsistency　remains　unsolved　in　particle　filter,　a　velocity　estimation　method　is　adopted　and　provides　robustly　converged　filter　results　as　no　error　accumulates　in　velocity　estimations.　To　reduce　computational　cost,　a　fast　laser　scanning　and　matching　implementation　is　proposed　to　facilitate　on　line　processing　in　real　time.　Furthermore,　all　particles　share　a　common　map,　which　allows　less　memory　and　computation　requirements.　Advantages　of　our　proposal　are　validated　by　both　simulations　and　real　experimental　results.