<p>Despite the rapid development of autonomous driving, drivers still need to take over when autonomous driving exceeds its design scope or malfunctions. The role of drivers is undergoing a significant transformation from operators to backup users. The existing human driving behaviour models focus more on the behaviour of humans as operators, with static and time-invariant characteristics. However, as backup users, human behaviour characteristics during the takeover process exhibit dynamic and time-varying characteristics, and traditional driver models can no longer describe these, making it difficult to support the safe development of autonomous driving. Unfortunately, the evolution mechanism of driver behaviour is unclear, which has led to the continuous occurrence of accidents in autonomous vehicles. To support the safe development of autonomous driving, we studied the changes in drivers’ cognition, decision-making, and control behaviours during the takeover, revealing the evolution mechanism of driver behaviours during the takeover. On this basis, a progressive reshaping model of human driving behaviours is constructed. The comparison with actual driver control data shows that the accuracy of the proposed model is 88.57%, providing a new perspective for understanding driver behaviour during emergency takeover and having certain application value in the research of autonomous driving technology.</p>