This paper addresses the issue of decreased speed prediction accuracy in tracked vehicles due to external noise during operation, and proposes an adaptive speed prediction method based on fuzzy logic. Traditional prediction methods based on physical models struggle to handle the complex dynamic characteristics unique to tracked vehicles, while GPS-based speed measurement methods have poor reliability in areas with signal obstructions. In this study, Hall sensors are used to collect real-time motor speed data, which is preprocessed through mean filtering and outlier removal, and a piecewise linear regression model is established. On this basis, the fitting parameters are dynamically adjusted using fuzzy logic. Experimental results show that during acceleration (0→1.2 m/s), deceleration (1.2→0 m/s), and constant speed (0.4/0.8/1.2 m/s) phases, the maximum absolute error of this method is less than 0.234 m/s (deviation less than 20%), and the standard deviation is all below 5% of the target speed. Under conditions of rapid speed changes, this method still maintains good prediction stability, verifying its application value and robustness in the special operating environment of tracked vehicles.