Deep Learning-Based Prediction and Optimized Path Planning for Sports Athletes’ Movement Trajectories

Abstract

In today’s increasingly popular sports, more and more scholars are focusing on the use of intelligent ways to further enhance the effectiveness of sports and competitions. In this paper, we use the LSTM network model to extract the sports characteristics of sports athletes, introduce the graph neural network GNN model to realize the spatial and temporal interaction design in the prediction of sports trajectory, and finally output the sports trajectory prediction model through the LSTM decoder module. Meanwhile, based on the trajectory prediction results and the optimization objectives and constraints of the sports trajectory, the sports path planning model is established using the quadratic planning model. The testing of the model found that its motion trajectory prediction model performance improved accuracy and EAO by 22.50% and 35.47%, respectively, compared to the baseline method. In addition, the model is able to accurately predict athletes’ movement trajectories in empirical analysis and can provide reasonable and optimal movement paths for soccer players. The model proposed in this paper can provide athletes with technical improvements and trajectory optimization solutions and formulate targeted training plans to strengthen theoretical guidance for sports athletes. The model can be improved and applied to real sports competitions in future research.

Keywords

• motion features
• lstm
• gnn
• trajectory prediction
• path planning
• 97p10