Optimal Dynamic Motion Sequence Generation for Multiple Harvesters

Abstract

Harvesting efficiency could be improved significantly by computing optimal harvesting patterns which minimize turning time. Hence, the execution of harvesting operations, especially in case of cooperating harvesters, needs to be carefully planned. In this paper, the operation planning for a fleet of harvesters is formulated as a discrete optimization, multi-Traveling Salesman Problem (m-TSP). Given number of “cities” and the cost traveling between every pair of them, the m-TSP searches for m round trips (one for every of m salesmen) in a way that every “city” is visited exactly once and the total cost is minimized. In our proposed formulation, the “cities” of the m- TSP correspond to the operating rows of the field. The cost is the nonproductive time, which is spent during the turnings at the headlands. This cost is computed based on the kinematics constraints of the vehicles and on the geometrical space constraints of the field. An existing heuristic algorithm with low computational requirements (in the order of a few seconds) was adopted for the solution of the m-TSP. The advantage of low computational time makes it feasible to re-plan an optimal fieldwork pattern when it is necessary for the remaining non-harvested field, while the harvesting procedure is being executed. Simulations of scenarios of planning and re-planning optimal harvesting operations are presented.