The goal of this project is to explore the potential of simulated annealing in solving a combinatorial optimization problem in agriculture. The specific problem addressed in this project is the generation of optimal crop rotation plans for farms.

The whole-farm planning problem is a practical challenge faced by the farmers and environmental and federal agencies. The goal of farm planning is to come up with crop rotation plans that are environmentally safe and economically feasible. The farmers are encouraged, and in some cases mandated, to follow National Resource Conservation Services (NRCS) guidelines in cropping and tillage practices in order to reduce the loss of top-soil due to water erosion, reduce the risk of pesticide and nutrient leaching, and reduce surface runoff risk. The conflicting nature of these goals with the farmers’ goals makes this a difficult problem and neither the agencies nor the farmers have the necessary tools to generate plans that satisfy the farmers’ need while adhering to environmental restrictions.

This project uses simulated annealing, a method derived from statistical mechanics, to find acceptable plans for farms with a wide variety of local and global constraints. We define the farm planning problem, lay the framework of the annealing algorithm, use the annealing algorithm to produce plans for a specific farm, compare the results with those obtained by another Al-based heuristic technique, and determine the plan’s feasibility in actual farming practice.

The framework upon which simulated annealing is based is very simple and most optimization problems can be formulated to fit this framework quite easily. Results obtained in this project are comparable with those obtained by another Al-based heuristic technique. Increasing the search space decreased the speed of convergence somewhat but for problems like farm planning where time is not extremely crucial, annealing seems to be a viable optimization tool.