Simultaneous　localization　and　mapping　(SLAM)　based　on　an　optimal　filter　is　widely　used　to　robot＇s　pose　estimation.　However,　this　type　of　SLAM　must　reconfigure　the　entire　state　vectors　in　every　iteration　as　the　number　of　landmarks　change,　which　results　in　the　state　vectors　needing　to　be　reconfigured　making　it　hard　to　isolate　potential　faults.　In　this　paper,　distributed　SLAM　based　on　advanced　distributed　unscented　particle　filter　(ADUPF)　is　presented　to　address　these　problems.　In　this　distributed　SLAM　system,　an　advanced　unscented　particle　filter　(AUPF)　was　used　in　every　local　filter　to　increase　the　accuracy　and　robustness.　In　the　ADUPF,　the　particle　distribution　inherited　initialization　algorithm　in　every　new　added　local　filter　was　proposed　to　improve　the　performance　of　a　random　initialization　algorithm.　The　experiment　results　show　that　the　distributed　SLAM　based　on　ADUPF　is　capable　of　improving　estimation　performance　compared　to　those　existing　methods.　©　2014　WIT　Press.