|M.Sc Student||Blass Tamir|
|Subject||Navigation Algorithm for Autonomous Vehicle Operating in|
|Department||Department of Civil and Environmental Engineering||Supervisors||Professor Raphael Linker|
|Professor Emeritus Itzhak Shmulevich|
|Full Thesis text - in Hebrew|
The technical developments of the last two decades provide a strong basis for the development of autonomous agricultural vehicles. So far such vehicles have been developed mostly for field agriculture, which typically include a small number of obstacles. By comparison, orchards include a large number of obstacles, and path-planning and navigation in such complex environments still presents many challenges.
The general goal of this research was to develop a path-planning algorithm for agricultural vehicles in orchards, assuming that the obstacle map is available. Such an assumption is realistic since most of the obstacles consist of trees, for which the planting map is available. The approach adopted in the present work is based on the “configuration space” method, which treats the vehicle as a point in the so-called configuration space that includes the representation of the obstacles for various vehicle headings. The path-planning algorithm itself combines two common algorithm, namely the A* algorithm and the visibility graph search. In order to make the path-planning procedure suitable for a vehicle operating in orchards, two main modifications are made: (1) the kinematics constraints of the vehicle (e.g. turning radius) are taken into account and (2) the path planning procedure takes into account user-defined agro-technical constraints. Such constraints are required in order to obtain realistic trajectories that avoid unsafe maneuvers and meet agricultural requirements such as avoiding unnecessary travel close to the trees or repeated travel over the same area. By comparison, existing algorithms based on the configuration space approach typically consider that the traveling surface is flat and that obstacle avoidance is the only constraint. Furthermore, these algorithms do not consider safety issues such as traveling at steep angles, or problems associated with low trafficability. All these issues are addressed in the path-planning procedure developed in the present work.
Various simulations that correspond to situations that are typical to orchard agriculture, such as end-of-row turning and safe travel to a designated tree, are presented. These simulations show that all the planned paths match the vehicle capabilities on the one hand and the agricultural constraints on the other hand. The vehicle’s parameters, the safety constraints, and other user-defined constraints can be easily adjusted to any specific application, which makes the system suitable for a wide range of vehicles and agricultural scenarios.