Abstract: Differential robots, which include both tracked and wheeled robots, have simplified structures and the capability for in situ steering when compared with other robot types, such as car-like robots. Because these features endow enhanced flexibility and scalability, differential robots have been widely used in intelligent manufacturing, logistics, military, agriculture, and other fields. Path tracking control is a pivotal technology within the autonomous navigation system of differential robots, and it maintains the differential robot’s traveling behavior along a given reference path by minimizing the distance between the robot and path. However, existing path tracking control methods are vulnerable to deviations stemming from the coupling between yaw rate constraints and travel speeds. Specifically, the pendulum angular velocity limits of differential robots become increasingly pronounced at higher speeds owing to this coupling, thereby impeding steering capabilities. This coupling leads to a deterioration of the accuracy of the path tracking control when the reference path curvature is large. To address this issue, we introduce a novel path tracking control method based on nonlinear model predictive control (NMPC). Initially, the coupling between the longitudinal travel speed and yaw rate constraints, which primarily arises from constraint conversion, is analyzed. The constraint range of the system can be fully leveraged by directly employing the left and right track line speeds as inputs, thereby averting under-constraint issues. Subsequently, a kinematic model is formulated using the track line speeds as inputs, and a nonlinear prediction model is constructed accordingly. An optimization objective function is then devised by leveraging the coupling between the longitudinal travel speed and reference path points. Thus, an NMPC-based active speed-regulating path tracking algorithm tailored for differential robots operating under constraint coupling is developed. Finally, an active speed-regulating path tracking control system is developed for differential robots to validate the efficacy of the proposed control method. The results of Simulink simulations and real-world differential robot experiments demonstrate that the proposed control system enhances the path tracking control accuracy of differential robots. Across all simulations and experiments, the absolute value of the maximum displacement error does not surpass 0.0723 m, whereas that of the heading error remains below 0.0964 rad. Compared with constant-speed path tracking control systems based on NMPC and linear model predictive control, the proposed system reduces the maximum displacement error by up to 99.22% and the maximum heading error by up to 93.32%. Furthermore, in contrast to an existing active speed-regulating path tracking control system that combines a speed adjustment controller with an NMPC path-tracking controller, the proposed system decreases the maximum displacement error magnitude by 87.55% and maximum heading error magnitude by 29.69%. Notably, the computation time of the control system does not exceed 18.00 ms throughout the simulations and experiments, with the control period set to a constant 50 ms. As the computation time of the proposed control system is significantly less than the control period, the system can satisfy the demand for path tracking control in real time.