Abstract: The shortage of agricultural workers and rising operational costs are increasing the demand for autonomous farm machinery. However, current technical solutions face several challenges, including dependence on regular-shaped fields, separation between path planning and velocity planning, and inadequate control accuracy in complex environments. This paper aimed to develop an integrated trajectory planning and tracking control method to enhance the operational performance of unmanned agricultural machinery under realistic field conditions. This paper proposed a cooperative optimization framework that integrated trajectory planning with tracking control. For the trajectory planning, the Fields2Cover library was employed to generate full-coverage paths for irregular fields, utilizing multiple objective functions, swath sorting methods, and turning connection strategies. To address the disconnection between path and velocity planning, an adaptive velocity switching strategy based on path curvature was introduced, combined with S-curve velocity profiling to ensure smooth velocity transitions. For tracking control, an improved Linear Quadratic Regulator (LQR) controller was designed based on an Ackermann steering kinematic model. The key innovation was a curvature-adaptive weight matrix adjustment mechanism that dynamically optimized the controller's parameters according to real-time path curvature. Additionally, a target steering angle preview feedforward strategy was implemented to reduce overshoot and enhance tracking precision during curves. Comprehensive simulation and real-world experiments demonstrated the effectiveness of the proposed method. In the simulation ablation study, the preview LQR reduced the mean and maximum lateral errors by 30.0% and 33.6% compared to traditional LQR, demonstrating the independent contribution of the preview mechanism in suppressing curve overshoot. Regarding overall performance, compared with traditional LQR, preview LQR, pure pursuit, and Stanley methods, the improved LQR reduced the mean lateral error by 50.0%, 28.6%, 21.1%, and 40.0%, and the maximum lateral error by 49.6%, 24.1%, 4.3%, and 59.8%, respectively. For heading error, the improved LQR achieved mean and maximum values of 0.005 rad and 0.064 rad, reducing the mean heading error by 28.6%, 16.7%, 16.7%, and 64.3%, and the maximum heading error by 31.9%, 28.1%, 29.7%, and 77.6% compared to the four benchmark methods. The velocity errors were also minimized at 0.014 m/s (mean) and 0.036 m/s (maximum). Real-vehicle tests on uneven grassland further validated the method's robustness. Quantitative analysis of the ablation group showed that compared to traditional LQR, the preview LQR reduced the mean and maximum lateral errors by 26.1% and 23.3%, and the mean and maximum heading error by 13.3% and 3.9%, effectively validating the mechanism's role in enhancing heading stability. Furthermore, compared to traditional LQR, preview LQR, pure pursuit, and Stanley methods, the improved LQR reduced the mean lateral error by 48.9%, 30.9%, 13.0%, and 56.5%, and the maximum lateral error by 43.9%, 26.9%, 22.8%, and 72.7%, respectively. The mean heading error reached 0.018 rad, representing reductions of 40.0%, 30.8%, 18.2%, and 71.0% compared to the four benchmark methods. Although the maximum heading error showed a 15.5% increase compared to pure pursuit, it decreased by 9.4%, 5.7% and 49.8% compared to traditional LQR, preview LQR, and Stanley. In velocity tracking, the improved LQR exhibited the smallest fluctuations, confirming superior stability under challenging field conditions. The integrated path-velocity-control cooperative optimization solution presented in this paper significantly enhanced tracking accuracy, operational smoothness, and field adaptability of unmanned agricultural machinery. This approach effectively addressed key limitations of traditional methods through synergistic planning and adaptive control. Therefore, this study can provide a solid technical foundation for the deployment of intelligent agricultural equipment in irregular terrains. Future research will explore reinforcement learning for automated parameter tuning to further improve system adaptability and deployment efficiency.