The article presents an intelligent automatic control algorithm for a small multi-rotor unmanned aerial vehicle (UAV) that implements real-time adaptation of a PID controller based on the assessment of object dynamics. The algorithm uses a first-order model approximation with online parameter estimation using the recursive least squares method and noise filtering using a Kalman filter. This approach provides computational efficiency and robustness to disturbances, which is critical for application on resource-constrained platforms. Algorithm tuning is performed taking into account constraints on coefficients and smoothing of estimates, which allows the system to function stably under changes in mass, external disturbances, and loss of GNSS signal. The proposed architecture can be integrated into existing control systems and expanded through the use of nonlinear models and adaptive positioning loops.