Abstract: Aiming at the problems that conventional deadbeat predictive current control (DPCC) strongly depends on motor parameters and the conventional extended state observer (ESO) with fixed input gain and bandwidth cannot simultaneously achieve parameter adaptability, disturbance tracking speed, and noise suppression performance, an artificial lemming algorithm (ALA)-optimized augmented extended state observer (AESO)-based model-free deadbeat predictive current control method for permanent magnet synchronous motors (PMSMs), namely ALA-AESO-MFDPCC, was proposed. First, the input gain was introduced as an extended state into the observer to achieve the simultaneous estimation of the input gain and lumped disturbance, thereby reducing the influence of inductance parameter variations on prediction accuracy. Second, the observer bandwidth was optimized online by the ALA according to the current tracking error, enabling the observer to improve disturbance estimation speed during dynamic operation and reduce noise amplification under steady-state conditions. Finally, the optimized observation information was incorporated into the model-free deadbeat predictive control law to achieve current prediction correction and disturbance compensation. Simulation results show that, compared with the conventional ESO-MFDPCC method, the proposed method reduces the total harmonic distortion (THD) by 41.0% and 40.7% under parameter-matched and parameter-mismatched conditions, respectively. Under the load torque step condition, the peak value of the q-axis current and THD are reduced by 31.6% and 55.9%, respectively. Furthermore, an RT-LAB hardware-in-the-loop experimental platform is established to verify the effectiveness of the proposed method under non-ideal conditions. The proposed method effectively reduces the dependence of conventional DPCC on motor parameters and improves the current control performance under parameter variations and load disturbances, providing an effective approach for enhancing the robustness of PMSM drive systems.