In　large-scale　internet-of-things　(IoT)　systems,　unmanned　aerial　vehicles　(UAV)　enabled　mobile　edge　computing　(MEC)　can　alleviate　the　performance　constraints　on　end　IoT　devices.　However,　due　to　the　uneven　distribution　of　IoT　devices　and　inefficient　problem-solving,　how　to　efficiently　perform　computation　offloading　in　large-scale　IoT　systems　is　a　major　challenge.　Existing　solutions　generally　cannot　fit　into　dynamic　multi-UAV　scenarios,　causing　inefficient　resource　utilization　and　excessive　response　delay.　To　address　these　important　challenges,　this　paper　proposes　a　novel　multi-UAV　deployment　and　collaborative　offloading　(MUCO)　method　for　large-scale　IoT　systems.　A　UAV　deployment　scheme　based　on　constrained　K-Means　clustering　is　designed　to　enhance　service　coverage　while　ensuring　balanced　coverage.　A　multi-UAV　collaborative　offloading　strategy　based　on　multi-agent　reinforcement　learning　(MARL)　is　developed　to　split　the　offloading　requests　from　IoT　devices　and　conduct　distributed　execution,　thereby　realizing　efficient　collaborative　offloading.　Extensive　simulation　experiments　validate　the　effectiveness　of　the　proposed　MUCO　method.　Compared　to　benchmark　methods,　the　MUCO　method　can　achieve　an　average　improvement　of　about　23.82%　and　28.13%　improvement　in　UAV　deployment　performance　in　different　scenarios,　and　can　achieve　lower　latency　and　energy　consumption.　©　2025　Acta　Simulata　Systematica　Sinica.　All　rights　reserved.