|M.Sc Student||Shani Michal|
|Subject||Trajectory Planning and Guidance of a Robotic Arm for Tree|
|Department||Department of Civil and Environmental Engineering||Supervisors||Professor Raphael Linker|
|Professor Emeritus Shaul Gutman|
|Full Thesis text - in Hebrew|
According to Ministry of Agriculture data, pruning and shaping of orchard trees represents more than 25% of the work-required in orchards. The Ministry of Agriculture decided upon the development of a mixed robotic-human operator system for pruning and shaping plantation trees. In this system, decisions regarding the location and type of pruning are determined by a human operator, while the operation the pruning tool itself is carried out automatically by the system.
The present research focused on some of the automated stages of the system, including:
• Planning of the trajectory according to pruning targets
• Guidance of the robotic arm along the trajectory, without damaging to the robot and branches
As a preliminary development stage of the system, it was decided to model the main branch as a spatial bent cylinder. This allowed us to use as starting point an existing algorithm for tracking cylinders, and adapt this algorithm to work with branches of deciduous trees. The tracking algorithm was based on the use of a camera positioned on the end-effector of the robotic arm and on geometrical data about the scanned cylinder. The control signal was calculated based on data retrieved from pictures received from the system camera in real-time. The main branch was identified via its edges, which appeared as two straight lines which were identified by Hough transform. The properties of these lines provided the feedback information based on which the control signals were calculated.
A mathematical correlation was found between the tracking speed and the tracking errors, and as a result, a method for calculating the maximal allowed speed was developed.
The tracking performance was evaluated for a straight cylinder, an arc-curved cylinder and a knee-curved cylinder typical of deciduous orchard trees. The tracking was carried out from various directions relative to the cylinder. These tracking tests were carried out at a maximum speed of 5 cm/s. It was found that for a straight cylinder, the task was carried out without any problems. For the curved cylinders, larger errors were received but the task was carried out successfully. The performance of the system was poorer when tracking a trajectory from the curved sides of the cylinder. This lower performance was due to a reduction in the validity of the assumption of a straight cylinder on which the tracking algorithm is based. As expected, it was found that reducing the tracking speed improved the results for these cases.