The　distributed　algorithm　based　on　consensus　is　commonly　used　to　track　a　moving　target　in　wireless　sensor　networks.　In　this　paper,　we　try　to　apply　it　to　cooperative　localization　by　analogizing　robots　as　nodes　in　a　multi-sensor　network　to　build　a　multi-robot　system,　robots　exchange　information　between　neighbours　only,　and　achieve　more　precise　localization　through　information　fusion.　And　we　detail　two　distributed　algorithms　based　on　consensus.　In　order　to　address　the　problem　of　low　precision　caused　by　the　time-varying　statistical　characteristics　of　the　noise　of　each　subsystem,　we　proposes　an　improved　distributed　state　estimation　based　on　consensus-estimation,　which　use　the　covariance　intersection　method　to　process　multi-source　data.　Finally,　we　build　a　simulation　platform　and　a　practicality　experiment　platform,　and　design　experiment　schemes　to　compare　the　advantages　and　disadvantages　of　these　algorithms.　©　2023　IEEE.