Due　to　the　dynamics　of　topology　and　resources,　Call　Admission　Control　(CAC)　plays　a　significant　role　for　increasing　resource　utilization　ratio　and　guaranteeing　users＇　QoS　requirements　in　wireless/mobile　networks.　In　this　paper,　a　dynamic　multi-threshold　CAC　scheme　is　proposed　to　serve　multi-class　service　in　a　wireless/mobile　network.　The　thresholds　are　renewed　at　the　beginning　of　each　time　interval　to　react　to　the　changing　mobility　rate　and　network　load.　To　find　suitable　thresholds,　a　reward-penalty　model　is　designed,　which　provides　different　priorities　between　different　service　classes　and　call　types　through　different　reward/penalty　policies　according　to　network　load　and　average　call　arrival　rate.　To　speed　up　the　running　time　of　CAC,　an　Optimized　Genetic　Algorithm　(OGA)　is　presented,　whose　components　such　as　encoding,　population　initialization,　fitness　function　and　mutation　etc.,　are　all　optimized　in　terms　of　the　traits　of　the　CAC　problem.　The　simulation　demonstrates　that　the　proposed　CAC　scheme　outperforms　the　similar　schemes,　which　means　the　optimization　is　realized.　Finally,　the　simulation　shows　the　efficiency　of　OGA.