In　this　paper,　we　develop　a　decentralized　tracking　control　(DTC)　strategy　to　stabilize　a　class　of　continuous-time　nonlinear　interconnected　large-scale　systems　in　the　presence　of　unmatched　external　disturbances.　This　strategy　is　derived　from　an　online　learning　approach　of　optimal　control.　Initially,　the　DTC　problem　is　formulated　to　ensure　both　tracking　control　of　isolated　subsystems　and　the　overall　stability　of　the　large-scale　system.　By　combining　the　tracking　error　with　the　reference　trajectory,　some　augmented　systems　are　constructed　and　then　the　DTC　design　is　transformed　into　an　optimal　control　problem.　Considering　external　disturbances,　we　formulated　it　as　a　two-player　zero-sum　game.　Critic　neural　networks　are　utilized　to　solve　the　Hamilton-Jacobi-Isaacs　equation.　This　approach　allows　for　the　estimation　of　Nash　equilibrium　solution　that　encompasses　both　the　optimal　control　law　and　the　worst-case　disturbance　law.　Notably,　a　novel　gradient　descent　strategy　with　momentum　is　established　to　tune　the　weights　of　the　critic　neural　network　for　approximating　the　cost　function　better.　Finally,　an　experimental　simulation　is　provided　to　verify　the　effectiveness　of　the　developed　DTC　scheme.