Abstract: To address the degradation in trajectory tracking and formation keeping caused by the high inertia and limited steering response of heavy duty unmanned mining trucks operating on open pit roads with spatially varying widths, a cooperative control method integrating diffeomorphic transformation and preview feedback is proposed. First, a kinematic model subject to nonholonomic constraints is established. A tangent based diffeomorphic transformation is employed to map the finite lateral safety region into an unbounded space, and a lateral control law is developed by integrating road width adaptive gains with preview heading feedback. Second, a cooperative longitudinal control law is constructed from the relative longitudinal position errors between adjacent trucks to maintain the desired formation spacing, while steering angle magnitude and rate constraints are incorporated to account for the physical response limitations of heavy duty steering actuators. The closed loop stability of the proposed method is subsequently analyzed through the longitudinal formation error system and the lateral tracking error system. Finally, numerical simulations are conducted on a composite road consisting of straight sections, continuous curves, and a progressive reduction in road width from 20 m to 10 m. A fixed operating point linear quadratic regulator is adopted as the benchmark controller. The results demonstrate stable cooperative operation of three unmanned mining trucks. The proposed method achieves a lateral tracking RMSE of 0.0757 m and an MAE of 0.1316 m, corresponding to reductions of 69.9% and 60.3%, respectively, relative to the benchmark controller. These results demonstrate that the proposed method improves lateral tracking accuracy and enhances adaptability to spatially varying road constraints, providing a new approach to cooperative control for heavy duty unmanned transport vehicles in open pit mines.