The　paradigm　example　of　social　interaction　is　cooperative　problem　solving,　in　which　a　group　of　autonomous　agents　choose　to　work　together　to　achieve　a　common　goal.　In　order　to　improve　existing　work　on　developing　formal　computing　model　for　cooperative　problem　solving　in　multi-agent　systems,　several　efforts　are　made　in　this　paper.　Firstly,　a　new　multi-agent　cooperation　logic　called　ATL-BDI　designed　in　our　previous　work　is　enhanced　so　as　to　make　it　more　suitable　for　modeling　cooperative　problem　solving.　Secondly,　basing　on　ATL-BDI,　the　cooperative　problem　solving　process　is　formally　described.　The　corresponding　new　computing　models　of　cooperation　recognition,　team　formation,　plan　formation　and　execution　are　developed.　The　iteration　of　these　steps　is　also　well　considered.　Several　important　properties　are　gained　and　proved　finally.