Autonomic　computing　is　expected　to　solve　system　management　complexity　and　cost　problems　in　IT　environment　by　enabling　systems　to　be　self-managing.　The　autonomic　computing　collaborative　work　based　on　policy-driven　is　studied,　including　competition　and　collaboration.　The　autonomic　elements　should　work　collaboratively　to　implement　the　complex　self-management　tasks　of　autonomic　computing　system.　The　competition　model　is　proposed　based　on　the　probability　policy,　and　the　equilibrium　solutions　are　studied　and　resolved.　It　can　improve　the　success　rate　and　utility　for　practical　applications　effectively.