When the tractor tows the tillage equipment for rotary tillage and ridge forming operations, uneven terrain and inconsistent soil compaction lead to unpredictable cultivation resistance, resulting in lateral torque that affects the vehicle’s travel posture and ultimately reduces the accuracy of navigation tracking. To achieve precise autonomous navigation for rotary tillage and ridge forming in complex agricultural environments, a control method was designed based on detecting lateral torque for autonomous navigation of the rotary tillage riding tractor. Firstly, based on the three-point hitch structure of the tractor and the operating scenario, the impact of forces at the tool suspension points on the navigation vehicle was analyzed. A two-dimensional axle pin sensor was developed to build a lateral torque collection system for detecting the lateral torque generated by the soil relative to the vehicle. The experiments demonstrated that lateral torque increased the lag in lateral errors during navigation operations. Secondly, the lateral moment of force was introduced into the kinematic model of the rotary tillage ridging tractor as an external random disturbance. A separation control strategy and linear feedback mechanism were used to calculate and predict state errors in real time, leading to the design of a robust model predictive control（ RMPC ）algorithm controller. Finally, a navigation control system was developed based on the ROS framework, and the software and hardware were integrated and deployed on a rotary tillage ridging tractor. Field operation experiments were conducted in a cabbage field during a single stubble treatment. The results indicated that the average absolute value of the lateral navigation error was 0.022 m, the average absolute value of the heading error was 0.034 rad, and the average linearity of the ridges was 4.4 cm, demonstrating that the overall operation quality met crop planting requirements.