Real-time positioning algorithm for variable-geometry air-assisted orchard sprayer

An algorithm for positioning of spraying arms based on laser scanner measurements for a variable-geometry air-assisted orchard sprayer is proposed. Functioning of the algorithm is presented for a real sprayer intended for experimental spraying in orchards, which has three hydraulically movable spraying arms that cover one side of a row of trees. The developed algorithm is based on realtime measurements of tree canopies from a laser scanner mounted on the sprayer. According to the measured canopies of trees the spraying arms move linearly and angularly in a plane perpendicular to the row so that spraying nozzles which are located at the ends of the arms are optimally positioned. The algorithm calculates the optimal position for each of the three height segments of a tree (each covered by one arm) based on a simplified contour of the measured canopy of the corresponding row section. Inside the height segments the contour is simplified with a linear approximation using the least-squares method. The optimal position for each arm is then calculated so that the nozzle is directed normally to the linear fit of the contour at a distance for which full coverage of the tree height segment is achieved. The coverage is calculated with consideration of the spray angle. Due to sometimes great displacements between positions in consecutive time steps, calculated positions in each time step may not always be reachable in a real operation of the sprayer. Such great displacements are mainly a result of variable growth of trees and angular sensitivity of the algorithm. The latter is expressed when a change in the contour shape causes a changed direction of a normal which further on causes a change in the position of the arm. This change of position can be interpreted in relation to the change of the contour shape as amplified proportionally to the distance between the arm and the linear fit. To obtain physically feasible displacements, calculated positions were smoothed using the unweighted moving average. The effect of moving average width is also described in the results. With more target-directed spraying it is expected that the drift and ground deposits of pesticides can be reduced. In addition, more effective spraying enables changes in the effective dose, resulting in smaller amounts of pesticides used.