Perception issues are one of the main causes of takeovers, with complex environments and sensor limitations often leading to system misjudgments. Prediction problems are also a key challenge, as autonomous driving systems perform poorly when dealing with the uncertain behaviors of pedestrians and vehicles.

Specific case analyses show that in scenarios such as left turns, right turns, lane changes, intersections, and crosswalks, autonomous vehicles frequently exhibit unreasonable braking and improper speed control. Additionally, unstable perception and unreasonable trajectory planning are also common reasons for takeovers.

Although modern perception technology has made significant progress, its performance in complex environments still needs improvement. Prediction and decision-making algorithms also need to be more stable and reliable when facing dynamic traffic environments. Regulatory and ethical issues are also major obstacles to the application of autonomous driving technology.