Urban　mining　recovery　and　reuse　involve　a　complex　system　of　multiple　agents　with　independent　interests　across　different　stages.　This　study　examines　the　dual-layer　coordination　among　four　key　agents—supplying,　recycling　and　sorting,　reuse,　and　regulatory—using　multi-period　game　theory　and　a　heterogeneous　social　network　model.　By　integrating　these　models　with　a　multi-agent　based　simulation　(MABS),　we　achieve　bidirectional　feedback　and　dynamic　interactions.　Our　analysis　under　various　policy　scenarios　reveals　that　non-cooperative　strategies　often　dominate,　but　adjusting　price　schemes　can　effectively　stimulate　cooperative　strategies　among　recycling　and　sorting　agents　with　limited　impact　on　other　agents.　Policy　effects　also　vary　based　on　agent　personality　matches.　A　gradual　subsidy　phase-out　mechanism　enhances　the　stability　of　collaborative　evolution　among　recycling　and　sorting　agents,　while　technical　threshold　subsidies　promote　technological　learning,　driving　refined　processing　by　reuse　agents.　Enhancing　network　effects　through　adjusting　interaction　willingness　and　market　entry　and　exit　rules　increases　network　link　probabilities　but　may　counteract　collaborative　behaviors,　whereas　small　groups　within　a　certain　range　show　better　coordination　and　stability.　In　this　study,　a　solid　basis　is　laid　for　future　research　and　implementation　of　effective　multi-agent　behaviors　in　urban　mining,　offering　a　valuable　reference　for　stakeholders　in　urban　resource　management　and　policy-making.　©　2025　The　Authors