Topic: edge computing for AR/VR game. 1. Apply ML based method to syncrohize pose tracking data that are recorded under different reference frame. (P2) 2. Comparison of prediction performance for different user motion patterns, between deep learning LSTM and control based Kalman filter. And adopt GNN to improve accuracy of ESKF based prediction considering dynamic horizon.(P3) 3. Explore multi-user motion prediction that runs concurrently in terms of accuracy and resource allocation (P4)