Analysis on picking plant sequence and route of transplanting robotic for column cultivation

Abstract: Column cultivation is a promising high-yield mode in protected agriculture which has been studied and developed for several years. However, its extreme difficulty of manual management has hindered its popularization and application. As a result, an automatic operation of seedling transplanting to column was put forward. A new-type spiral cultivation column and matched seedling transplanting robot that was composed of rail-mounted autonomous vehicle, slide-arm type manipulator, double-needle electromagnet-driven end-effector, and a turning seedling supply device was developed. In view of the frequently repetitive operation of robotic seedling transplanting with changing starting and ending points of routes, the optimal design of a pick-plant sequence and non-collision shortest routing are essential to improving the success rate and efficiency. In robotic seedling transplanting, the route of a manipulator's ending is usually composed of picking an up-down route, a plant up-down route, and a transport route from pick top-point to plant top-point. For column cultivation, the transport route is quite different with regular robotic transplanting in view of its three-dimensional feature. So in this paper, a design of the shortest non-collision route was performed first for robotic seedling transplanting to spiral cultivation columns. In optimal route design, different principles were adopted for different relationships between the height of the plug position and height of the plant position in a column, together with the issue of seedling avoidance since a pick up-down route had to be extended to avoid a collision with other seedlings. And then the shortest non-collision route length of thirty-two different sequential pick-plant plans and a plan by a nearest neighbor algorithm were compared. Sequential pick plans were divided into S-shape and equidirectional seedling pick sequences that had different starting points, while sequential plant plans were divided into bottom-to-top and top-to-bottom sequences. The study indicated that the difference between the maximal and the minimal accumulated route length was more than 25% when seedlings of a total plug were transplanted to several one-meter-high columns. The pick sequence had an important effect on route length, which may mainly be attributed to the necessary of seedling avoidance. The computed results indicated that thirty-two plans were distributed in three areas, and each area had similar seedling avoidance times and accumulated route lengths. The plan by the nearest neighbor algorithm and equidirectional sequential pick with bottom-to-top plant had a stable value of route length when one plug of seedlings was transplanted to multiple columns. The effect of plant sequence on route length obviously rose with the increase of column height, and the average accumulated route length of 16 pick sequences when transplanting to three-meter-high column from top to bottom was 9.83% higher than transplanting from bottom to top. We found that the steady optimal schemes were pick-plant sequence searched by nearest neighbor algorithm and equidirectional sequential pick with bottom-to-top plant. To apply in control of robotic seedling transplanting, a plan of equidirectional sequential pick with bottom-to-top plant may be selected in priority in view of its obvious advantage of real-time rapid execution.