Due to the characteristics of inductance parameter mismatch and back electromotive force harmonics caused by novel leakage flux branches and other non-ideal factors for the variable-leakage-flux permanent magnet (VLF-PM) motor, its control system suffers from a deteriorated performance of the rotor position estimation. To overcome the problems mentioned above, an adaptive tracking estimator of the rotor position is proposed in this paper for the VLF-PM motor control system. First, the proposed method simplifies the VLF-PM motor mathematical model and reduces the effect of inductance parameter variations according to the active flux concept. Then, robust and gradient descent algorithms are utilized to maintain the robustness of inductance parameter variations and eliminate the specific order harmonics owing to the novel leakage flux branches. Meanwhile, the accuracy and stability are enhanced. Furthermore, the position compensation based on the current adaptive tracking strategy is proposed to compensate the rotor position error caused by other non-ideal factors. Finally, the feasibility of the proposed estimated system is verified.