Abstract: To mimic the human steering manoeuvre in curve driving, a human-like driver model with visual perception and decision-making modules was proposed. Inspired by the studies of human visual behavior in curve driving, the visual perception module was established to extract the visual inputs of vehicle speed, lateral deviation at near zone and heading angle error at far zone for steering decision from road information and vehicle motion states. Because of the fuzzy reasoning and complex characteristic of human driving behavior, the driving simulator experiments were conducted on multi-curvature curved roads at different speeds to acquire the training data, and an adaptive neuro-fuzzy inference system（ANFIS） was adopted to establish the decision-making module of steering wheel angle. Based on PreScan/Simulink joint simulation, the human-like driver model was verified at different speeds on different curved roads. The results show that the model has basic road following ability, and the generated angle changes smoothly, which can reflect the general human steering behavior in curve driving with a certain applicability to other driving conditions.