Environmental　perception　is　one　of　the　key　issues　in　robot　research.　To　complete　more　precise　tasks,　robots　need　a　recognition　ability　to　recognize　all　dynamic　and　static　objects　in　the　environment　accurately.　Currently,　the　prevailing　approach　in　robotics　is　to　employ　object　detection　methods　for　perceiving　stationary　objects　in　the　robot＇s　vicinity.　This　primarily　entails　the　use　of　2D　object　detection　based　on　images　and　3D　object　detection　based　on　depth　images　or　point　cloud　data.　From　an　input　data　perspective,　it　is　evident　that　depth　images　and　point　cloud　data　offer　a　substantially　richer　set　of　three-dimensional　spatial　information　compared　to　conventional　image　data.　Therefore,　3D　object　detection　is　more　suitable　for　robot　environmental　perception.　This　article　provides　an　overview　of　3D　object　detection　methods　for　facial　robots.　This　article　mainly　reviews　methods　based　on　radar　point　cloud　information,　introduces　the　historical　evolution　of　3D　object　detection,　and　elaborates　on　the　performance　and　limitations　of　different　methods.　At　the　same　time,　it　lists　commonly　used　datasets　for　related　tasks.　©　The　Author(s),　under　exclusive　license　to　Springer　Nature　Singapore　Pte　Ltd.　2024.