Research on Autonomous Navigation and In-Forest Path Decision-Making Mechanisms for Rubber Tapping Robots

Natural rubber holds a strategically important position in the national economy, and its production still relies primarily on manual rubber tapping. However, manual tapping is physically demanding and offers relatively low economic returns, which has led to an aging workforce and a shortage of young labor. As a result, the natural rubber industry is facing a severe labor crisis, making the intelligent upgrading of rubber tapping operations an inevitable trend. Robot navigation in rubber plantations often encounters a range of environmental and operational challenges. Therefore, to address autonomous in-plantation navigation for rubber tapping robots, this study aims to develop an autonomous navigation system and an in-forest path decision-making mechanism based on multi-objective optimization. In response to the practical requirements of rubber tapping operations, the hardware platform of the system was selected and assembled, including a LiDAR sensor, an IMU (Inertial Measurement Unit), and the main controller. In parallel, the software platform was established by setting up the ROS (Robot Operating System) development environment and designing navigation-related software modules, thereby constructing the integrated hardware–software platform for autonomous navigation of a rubber tapping robot. On the premise of time synchronization across sensor streams, algorithms such as the Kalman filter, complementary filter, and extended Kalman filter were employed to fuse data from the LiDAR, IMU, and wheel odometry. Combined with the Cartographer algorithm and its parameter tuning, the system achieved environmental perception and accurate localization. Robot path planning was then realized through Dijkstra global planning and the DWA (Dynamic Window Approach) local planning algorithm, enabling full navigation capability. Furthermore, a multi-objective-optimization-based in-forest path decision-making mechanism was proposed, which includes acquiring target-point coordinates, selecting the optimal coordinates, generating additional coordinates, and optimizing the path. This mechanism allows the robot to autonomously determine the optimal navigation sequence regardless of its starting position on the map, thereby improving system autonomy and enhancing the rationality of both the navigation order and the planned routes, ultimately addressing multi-target navigation in forest environments. Finally, field experiments were conducted in a rubber plantation to validate the proposed autonomous navigation system and the multi-objective-optimization-based in-forest path decision-making mechanism, and the results were analyzed. The experimental results indicate that, when the acceptance radius of the autonomous navigation system was set to 0.3 m, the mean navigation error was 0.29103 m, the root mean square error was 0.29999 m, and the variance was 0.00529568. The maximum distance between the robot center and the tree was 0.65622 m, which falls within the effective operating radius of most robotic manipulators. The overall rationality of the planned routes generated by the in-forest path decision-making mechanism was 92.14%. These results meet the navigation requirements for rubber tapping operations.