In order to solve the problems of large differences in task types, strong fluctuations in resource states, and difficulty in balancing delay and utilization in static scheduling in heterogeneous computing environments, this paper proposes an adaptive resource scheduling algorithm and a dynamic resource decision-making model. Based on the unified representation of task-resource, this method integrates heterogeneous perception initialization, nonlinear search control and feedback-driven update into the same scheduling link, so that task mapping, resource reconfiguration and parameter correction can be continuously adjusted with the change of system state. Based on the extended CloudSim platform, comparative experiments were carried out on HPC2N, NASA and synthetic burst workloads. The results show that under the condition of 2000 tasks, the makespan of the proposed method is reduced to 1214 s, which is further decreased by 8.4% compared with the optimal comparison method. The resource utilization rate is increased to 85.6%, and the average response delay is controlled at 109 ms. Research shows that the model can effectively alleviate the node imbalance and task congestion problems in heterogeneous resource pools, and has good adaptability and practical value for online resource scheduling in complex computing scenarios.
