| China is the country with the large citrus planting and production in the world.The manual harvesting mode has the outstanding problems of high production cost and high labor intensity.With the aging population,the citrus harvesting robots have become a real need for agricultural modernization.In the natural environment,the position of citrus fruits is random,and the distribution of branches,leaves and other obstacles is chaotic,which brings great difficulties to the movement planning of the citrus harvesting manipulator.In order to make harvesting manipulator reach the specified position accurately without damaging the fruits and trees,the motion planning and obstacle avoidance algorithms of the harvesting manipulator need to be studied.In this paper,the motion planning algorithm of citrus harvesting manipulator in unstructured environment is studied.Based on the asymptotically optimal InformedRRT* algorithm for the citrus harvesting environment,PGI-RRT* algorithm is proposed,and it is verified by simulation environment and indoor and outdoor environment harvesting motion planning experiments.The main research contents and conclusions are as follows.(1)Through the study of citrus orchard harvesting environment,the harvesting coordinate system and the harvesting plane are defined and the obstacles around the fruit are described.The distribution model of citrus fruits and obstacles in the unstructured environment was summarized: there are branch obstacles in multiple directions around citrus fruits,and the average distance between obstacles and fruits in the harvesting plane is about 15 cm,and the average distance beyond the harvesting plane is about 14 cm.A simulation test platform of ROS-based harvesting robot was constructed to realize the reconstruction of citrus harvesting environment.(2)The existing common motion planning algorithm was studied.The simulation harvesting experiments were conducted using the commonly used manipulator motion planning algorithm based on random sampling,and the results showed that the existing commonly algorithms have difficulty in motion planning and low success rate in the un-structural environment.A two-stage harvesting motion planning method based on a harvesting guide point is proposed,and the verification through simulation experiments shows that the harvesting method can effectively improve the success rate of harvesting manipulator motion planning in unstructural environment.Then PGI-RRT* algorithm is proposed,which introduces pre-harvesting guide point in the configuration space,realizes parallel expansion of four random trees between the initial point,pre-harvesting guide point and target point,introduces P-probability sampling,adds adaptive dynamic step size adjustment function,and introduces heuristic node optimization strategy function.(3)The motion planning experiments of the PGI-RRT* algorithm are conducted in 2D and 3D simulation environments.The 2D simulation planning experiments of the algorithm show that the PGI-RRT* algorithm reduces the time to find the initial path by about 75%-86% and increases the success rate by 22%-32% compared to other motion planning algorithms.The programming implementation of the PGI-RRT*algorithm is implemented on ROS.The effect of different horizontal pre-harvesting guide point location settings on planning success and path cost was investigated and the optimal horizontal pre-harvesting guide point location was obtained.the PGI-RRT*algorithm was validated in eight different citrus harvesting simulation environments.The test results show that PGI-RRT* can effectively complete the motion planning of the harvesting manipulator,and the planning success rate is improved to 96%.(4)Indoor and outdoor citrus harvesting motion planning experiments were conducted by combining a real manipulator and a vision system.The experiments show that the PGI-RRT* algorithm can significantly reduce the planning time and improve the planning success rate.In the outdoor orchard environment,the planning success rate of the RRT-connected algorithm is only 28%,while the overall success rate of the PGIRRT* algorithm is 95%;compared with the RRT-connected and Informed-RRT*algorithms,the PGI-RRT* algorithm can reduce the motion planning time by 40%-60%,improve the planning success rate by 6-15%,and shorten the manipulator’s execution time by about 8s. |
PDF Full Text Request