In　recent　years,　large-scale　events　have　been　held　frequently,　and　more　and　more　people　are　participating　in　these　activities.　When　encountering　an　emergency,　it　is　necessary　to　quickly　evacuate　the　participating　people.　A　reasonable　traffic　evacuation　plan　is　an　important　part　of　the　efficient　evacuation.　However,　the　uneven　geographical　distribution　of　a　large　number　of　vehicles　and　participants　poses　a　challenge　for　efficient　evacuation　planning.　An　efficient　emergency　evacuation　plan　needs　to　minimise　evacuation　time　and　traffic　congestion.　Reasonable　route　selection　and　traffic　flow　assessment　are　key　to　the　evacuation　plan.　In　this　paper,　we　propose　an　efficient　multi-vehicle　navigation　method　based　on　trajectory　vector　features　considering　non-uniform　destination　distribution　for　emergency　evacuation.　First,　we　employ　the　CrossRank　algorithm　to　extract　the　real-time　state　vector　of　roads　and　then　navigate　the　vehicle　based　on　these　state　vectors.　In　addition,　spatial　diversity　theory　is　introduced　into　our　model　for　the　non-uniform　distribution　of　multiple　vehicle　evacuation.　We　conduct　a　series　of　multi-vehicle　navigation　simulation　experiment　on　a　real　taxi　trajectory　data　set.　The　experimental　results　demonstrate　that　our　approach　is　effective　and　efficient.　Copyright　©　2019　Inderscience　Enterprises　Ltd.