This　paper　proposes　an　optimal　output　synchronization　control　method　for　heterogeneous　multi　-agent　systems　(HMASs)　under　a　performance　-dependent　switching　topology　and　DoS　attacks.　First,　local　and　global　switched　performance　index　functions　(SPIF)　and　SPIF-dependent　topology　switching　law　are　proposed,　respectively,　thus,　the　control　performance　and　topology　quality　can　be　quantitatively　expressed.　Second,　an　adaptive　dynamic　programming　(ADP)　algorithm　with　mode　switching　is　proposed,　aimed　at　dealing　with　the　difficult　Hamilton-Jacobi-　Bellman　equation,　as　well　as　the　analytical　complexity　caused　by　the　switching　dynamics.　The　convergence　of　the　switched　ADP　algorithm　is　proven　to　ensure　its　correct　implementation.　Then,　for　different　topologies,　multi　-mode　Actor-Critic　neural　networks　(NNs)　are　built　for　each　agent　to　calculate　optimized　control　policies　and　SPIF,　respectively.　Furthermore,　an　NNbased　state　compensation　mechanism　is　designed　to　expand　the　applicability　of　the　designed　switched　ADP　algorithm　when　the　leader＇s　output　transmission　is　unreliable.　Finally,　the　results　of　numerical　examples　confirm　that　the　proposed　method　is　feasible.