Abstract: To address the issues of weak adaptive capability of impedance parameters and insufficient anti-disturbance ability in robot constant force control under uncertain environments, we propose a compliant control method integrating fractional order active disturbance rejection control (FOADRC) and radial basis function neural network (RBFNN). Firstly, we design a FOADRC composite controller in the inner loop of position control to estimate and compensate the total system disturbance in real time. Secondly, we construct an adaptive variable impedance model based on RBFNN in the outer loop, where convolutional operations are embedded in the hidden layer of the neural network to accelerate feature extraction, realizing the nonlinear mapping from force/position errors to impedance parameters. Furthermore, we use the improved artificial lemming algorithm (IALA) to tune the controller parameters in both the inner and outer loops. Simulation experiments show that the proposed method can better achieve the desired force tracking, reduce the force overshoot and force fluctuations during contact.