An adaptive algorithm for compensating for the accumulation of errors in a platform-free inertial navigation system designed for autonomous navigation of unmanned aerial vehicles has been proposed. Numerical simulations of a three-axis INS with typical accelerometer and gyroscope errors, including random noise, zero instability, and drift, have been conducted. The algorithm implements the adjustment of error model covariances based on the analysis of the Kalman filter prediction error, ensuring stable suppression of accumulated errors without external correction.