To diminish the position estimation error with harmonic fluctuations for position sensorless interior permanent-magnet synchronous motor (IPMSM) drives, a multiple adaptive vector filter (AVF) cross-feedback network (CFN) is proposed for a model-based position observer. A full-order Luenberger observer is utilized to obtain back electromotive force (EMF) information. The harmonic fluctuations in the position estimation error caused by inverter nonlinearities and flux spatial harmonics are analyzed. Accordingly, a newly proposed multiple-AVF CFN with an easy-to-implement structure is adopted to detect and compensate the back-EMF harmonic distortions and, hence, to eliminate the position error harmonic fluctuations. Consequently, the position estimation accuracy can be enhanced, which improves the position sensorless control performance. The proposed scheme was verified on a 2.2-kW sensorless IPMSM drive. Experimental results confirm the effectiveness of the method in eliminating the harmonic fluctuations in the position estimation error.