Under the background of digital transformation of national fitness public service, traditional manual scheduling and regular threshold scheduling are difficult to adapt to the fluctuation of fitness demand period, heterogeneous distribution of resources and changes in user feedback. For public sports venues, intelligent fitness equipment areas and community service areas, this paper constructs a dynamic scheduling and service optimization model based on PPO reinforcement learning. Multi-source data perception, time series demand prediction, resource load estimation and multi-objective reward function are used to realize the collaborative optimization of resource allocation, time period adjustment and user guidance. The model converts the reservation record, passenger flow monitoring, equipment status, external environment and user evaluation into computable state input, and improves the iterative stability of the strategy through feedback reward correction. The experimental results show that the average relative error of the model for the demand prediction of four types of fitness projects is 5.7%, and the accuracy of resource load level recognition reaches 92.7%. Compared with manual experience scheduling, PPO scheduling improves resource utilization to 89.3%, reduces average response time to 29.4 s, and user satisfaction reaches 91.8%. The research can provide technical support for the intelligent scheduling and fine governance of national fitness public services.
