Path planning algorithms for agricultural field machines

OF DOCTORAL DISSERTATION HELSINKI UNIVERSITY OF TECHNOLOGY P.O. BOX 1000, FI-02015 TKK http://www.tkk.fi Author Timo Oksanen Name of the dissertation Path Planning Algorithms for Agricultural Field Machines Manuscript submitted 23.8.2007 Manuscript revised Date of the defence 14.12.2007 Monograph Article dissertation (summary + original articles) Department Automation and Systems Technology Laboratory Automation Technology Laboratory Field of research Automation technology Opponent(s) Associate Research Professor Alonzo Kelly Supervisor Professor Arto Visala Instructor Abstract In this thesis, a coverage path planning problem is discussed in the case of agricultural fields and agricultural machines. Methods and algorithms to solve this problem are developed. The necessary condition is to cover the whole field, while the goal is to find as efficient a route as possible. As yet, there is no universal algorithm or method capable of solving the problem in all cases. In this thesis, two new approaches to solve the coverage path planning problem in the case of agricultural fields and agricultural machines are presented. In the first algorithm, the field plot is split into subfields that are simple to drive. Each subfield is constructed from trapezoids, the sides of which are attached. Thus the opposite sides of each subfield are parallel. In the search method, the best driving direction for each subfield is found. In the second algorithm, the path is planned on the basis of the machine's current state. Inspired by model predictive control, all the possible routes are simulated over the horizon of one route around the field or one back-and-forth swath. The best of the simulated routes is selected and the first part of that route is applied, up to the next turning. In the next turning, the algorithm is repeated. There are advantages and drawbacks in both algorithms. Neither of them solves the universal problem optimally. Nevertheless, the developed algorithms are remarkable steps towards finding a way to solve the problem. As a side-result, the properties of Finnish field plots are discussed and the field plot shape analysis is made. The classification of field plots based on shape was investigated, but only 25% of field plots belong to one or other simpleshape classes, such as rectangles and triangles, while the rest are more or less undetermined shapes. The other side-result is a trajectory generation for a tractor-trailer vehicle in a headland. The problem is formulated as an optimal control problem and solved using available methods and tools. With this approach, the trajectory can be solved for any headland width and angle.