Tractors have been widely used for production in modern agriculture. But the tractor rollover can be a frequent accident in agricultural production at present. It is still lacking in the reliable theoretical and technical support to the tractor rollover, because of the abruptness and rapidity of the accident. In this review, the research achievements were firstly summarized on the mathematical description of tractor kinematics and rollover dynamics. A systematic analysis was made on the strengths and weaknesses of dynamic and kinematic modeling approaches on the rollover characteristics of tractors. According to the variability of field operation environments and the complexity of the tractor rollover, machine learning was also introduced to treat the complex force relationships and unknown disturbances during tractor operation. Secondly, the research progresses were reviewed on the tractor rollover mechanisms, in order to extract the influencing factors of rollover. Furthermore, it was recommended that the influencing factors of rollover should be classified and quantified to explore the specific mechanism under different rollover conditions. It was necessary to establish the complete characterization of the influencing factors of rollover. Thirdly, the design principles of rollover warning were incorporated with the active anti-rollover control systems. The rollover risks were systematically summarized to assess the anti-rollover systems. Tractor parameters also shared a significant impact on the assessment indicators of rollover risk. The most effective solutions were using cost-saving intelligent sensors and soft computing with machine learning. The active anti-rollover control system was significantly improved to solve the insufficient anti-rollover torque and control timeliness in the active anti-rollover control system. More control schemes should be proposed to combine the multiple rollover prevention actuators and the pre-perception of the environment. Fourthly, the summary was also made on the theoretical basis and technical achievements of rollover protection devices. A pre-regulatable actuator should be designed on the intelligent, automated, or foldable rollover protection device in the active anti-rollover control system. Lastly, the experimental means were summarized for the rollover mechanism analysis and theoretical verification. The platforms should be recommended with the function of data process and posture control. The hardware simulation platforms were also integrated with the different simulation software. The reason was that the current lateral stability platforms were unable to reconstruct the complexity environment, and then represent the complexity of the tractor variation in real time. Research priorities and directions of tractor anti-rollover and rollover protection were specifically predicted to provide basic research and product development in modern agriculture.