The current algorithm of differential vehicle navigation cannot fully meet the optimal turning radius, long-distance path tracking, and simple navigation logic. In this study, a navigation control algorithm was proposed using ROS (robot operating system), according to modular and hierarchical development. The four-wheel differential speed vehicle was also transformed for testing. The target path was described by the target point plus the target yaw angle. The target path was an infinite straight line through the target point, whose direction was the target yaw angle. The vehicle started from any position, then approached the target path, and finally reached the target point. The kinematic equation of the vehicle was obtained by the kinematic analysis of the four-wheel differential speed vehicle. There was a relationship between the forward speed and steering speed of the vehicle, with the speed of the left and right wheels. Then the coordinate transformation was implemented to obtain the lateral deviation between the vehicle and the target path, and the longitudinal distance between the target point and the vehicle. The required forward speed was calculated using the longitudinal distance. The driving angle close to the target path was determined using the lateral deviation. Then the required steering speed was obtained after the difference between the driving angle and the yaw angle of the vehicle. Two advantages were gained during optimization. One was to improve the robustness and flexibility of the algorithm, which was stably solved under any size of lateral deviation and simplified the complexity of path planning. The algorithm was freely switched to the target path of the task, whether to focus or ignore the path planning among lines. Another was that the internal parameters of the algorithm all shared specific meanings, which were easy to understand and adjust, according to different needs under the different environments and operations. The above calculation was realized by the double variable limiter PID algorithm. The forward speed and steering speed were converted into the left and right wheel speeds through the kinematic equation of vehicles. The wheel speed was input into the closed-loop PID of the lower computer, in order to realize the accurate control of the left and right wheel speed. As such, the vehicle was stably controlled to reduce the disturbance during operation, according to the calculation of the algorithm. The accuracy and adaptability of the algorithm were then verified in three aspects: the average value of lateral deviation, The root mean square error, and the maximum deviation. The simulation results show that the average error was 1.146 cm, The root mean square error was 1.27 cm, and the line distance was less than 4.092 m when the four-wheel differential vehicle controlled by the path following algorithm runs at the speed of 0.4 m/s. The navigation system performed the better stability and adjustment. At the same time, the optimal driving trajectory was determined as the driving angle close to the target path, according to the unique lateral deviation of the algorithm. As such, the vehicle was used to drive perpendicular to the target path, whether far from or approaching the target path. The failure of the traditional algorithm was avoided by slowly adjusting the direction along the target path in the same environment. A comparison was then made on the different parameters in the simulation. The parameter adjustment of the algorithm was then verified to evaluate the influence of different parameters on the control system. In the cement ground test, the average error was 2.833 cm and the root mean square error was 3.21 cm when the vehicle was driving at the speed of 0.4 m/s on the cement pavement environment. The average error was 5.249 cm and the root mean square error was 6.13 cm when the vehicle was driving at the speed of 0.4 m/s on the fruit tree pavement environment in the field test. The average error rose slightly, and the variation coefficient changed little, with the increase of the ruggedness of the ground environment, indicating the strong robustness and the excellent anti-interference of the algorithm. The navigation system and the test vehicle shared better adaptability in different terrain environments. The finding can also provide a strong reference for the navigation technology of differential steering chassis.

Unmanned agricultural equipment can effectively solve the shortage of labor force in recent years. Environmental perception has been one of the most important components to realize the safe and reliable operation of intelligent agricultural machinery in complex unstructured farmland. However, it is still limited in the environmental perception, due to the dynamic and unstructured environment. Some challenges also remained to enhance the autonomous cognition and coordination of unmanned agricultural machinery in the operations of "plowing, planting, managing, and harvesting". The perception sensors are also required to realize robust and real-time environmental safety. In this study, the stable and reliable technologies of environment information perception were reviewed to improve the autonomous navigation and operation performance of unmanned agricultural machinery. Firstly, the typical elements of unmanned agricultural machinery were introduced into the operating environment and various types of sensors. Then, the information perception technologies were summarized from the aspects of the obstacle, crop line, farmland boundary, and elevation information perception. Finally, the challenges and trends were summarized from the perceptual performance, open access dataset, multi-sensor data fusion, data standardization, and interface standard. The complex, changeable and unstructured environment was highly required for the robustness and real-time performance of the sensing system in the unmanned working. It was still lacking in the unified benchmarks of environmental perception and performance evaluation standards. New technologies were expected to overcome the current challenges. The finding can also provide strong support to the modernization and intelligence of agricultural production.

To enhance adaptability in orchards with taller average tree heights and improve spraying effectiveness on Jinggang pomelo trees, this paper proposes a UAV-UGV cooperative targeted spraying system (UCTSS) and develops a prototype. The UCTSS primarily consists of a UAV and a UGV, networked using the Robot Operating System (ROS). During operation, both the UAV and UGV navigate between tree rows while carrying the spraying module. When the UAV reaches suitable spraying positions, the UGV halts to activate the spraying module, and the UAV performs targeted spraying from top to bottom. The paper employs a master-slave method for basic formation control of the UAV and UGV, resulting in an average tracking error of 0.118 m and a standard deviation of 0.040 m during testing. Additionally, a LiDAR-based targeted spraying detection method is designed and validated through simulation experiments, achieving an accuracy rate of 96% with an average position error of 0.13 m. Field trials in orchards demonstrate that the UCTSS meets stability requirements, with the average tracking error of the UAV measuring 0.158 m during coordinated movement and 0.013 m during spraying. In terms of spraying effectiv eness, the UCTSS exhibits higher average droplet density and deposition values at various heights of the same tree compared to the DJI-T50, along with a lower coefficient of variation between levels, resulting in a more uniform spraying effect. The feasibility of the UCTSS is validated, providing a novel approach for orchard protection in areas with taller average tree heights.