Abstract: The performance of task scheduling algorithm on heterogeneous MPSoC is affected by heterogeneous cores, run-time load and tasks dependencies. A novel load-aware task scheduling algorithm is proposed on heterogeneous MPSoC, which divides task-set into task-subsets based on tasks dependencies, computation overhead and communication overhead. In considering the core's load state, task-subsets are dispatched to appropriate cores by maximum weight matching of weighted bipartite graph, which improves the overall efficiency of task-set. Simulation results show that the proposed algorithm can reduce the length of task-set scheduling and improve the utilization of cores.