This study introduces an adaptive polynomial predictive filter (APPF) to address the issues of inconsistent and interrupted sensor data in estimation and prediction tasks. The APPF utilizes a polynomial prediction model to estimate time‐variant/time‐invariant sensor data through polynomial extrapolation. A key feature of the APPF is the autonomous determination of an accurate polynomial order, which reflects the nonlinearity level of the system. This dynamic adjustment allows the APPF to effectively capture the complex dynamics of nonlinear systems without manual tuning, thereby ensuring an accurate sensor data estimation and fusion. This study primarily focuses on validating the feasibility of the APPF method, and the theoretical basis, implementation, and performance evaluation of the APPF are covered, with results from MATLAB simulations and laboratory‐based practical experiments comparing the APPF with Kalman filter techniques. Experiments using Hall effect sensors, which provide precise localization by detecting magnetic signals, were conducted to assess the effectiveness of the APPF in addressing the sensor data variances and disturbances commonly encountered in surgical robotic tracking. The findings demonstrate that the APPF significantly enhances estimation and prediction accuracy, highlighting its potential for improving sensor data reliability in various applications.