<![CDATA[This study develops a Physics-Informed K-Nearest Neighbors (KNN) approach to reconstruct projectile motion based on experimental data using Tracker video analysis and Python. This study is motivated by the limitations of idealized physical models in representing real experimental conditions, which are influenced by air resistance, measurement noise, and data uncertainty. The research method comprises three main stages: acquisition of parabolic motion data using Tracker, theoretical modeling based on kinematic equations, and trajectory reconstruction using KNN. The results show that Physics-Informed KNN produces trajectories that are closer to the observed data compared to pure physics models. Horizontal motion exhibits the characteristics of uniform linear motion (ULM), while vertical motion forms a parabolic pattern consistent with uniformly accelerated linear motion (UALM). Error analysis indicates that the best performance is achieved at K = 2 with the minimum RMSE value. Based on the experimental dataset analyzed, the integration of machine learning and physical principles shows potential for improving the accuracy of dynamic system reconstruction under realistic experimental conditions.]]>
Copyrights © 2026
