Decision support system for sensor-based autonomous filling of grain containers

摘要

Autonomous technology in agriculture offers many products that reduce distractions and fatigue experienced by machinery operators, including automatic path guidance, variable rate product delivery, and precision seed placement. However, the size and complexity of modern mechanical harvesting operations have limited the ability of autonomous technology to significantly reduce total negative effects on grain combine operators. Combine operators are highly susceptible to fatigue because several tasks must be performed simultaneously to ensure safe machine operation. These duties include monitoring internal threshing and crop flow intake, maintaining row alignment, avoiding foreign material intake, and overseeing unloading grain.The primary goal of this project was to design a decision support system for autonomous unloading of combines. When unloading grain on-the-go, operators divert more attention away from critical tasks to focus on grain delivery to the adjacent cart. An autonomous system eliminating the need for combine operators to focus on unloading on-the-go potentially reduces operator stress and grain spillage.Critical to the decision support system for autonomous unloading was the input of a two-dimensional fill grid used to describe the grain height in the cart. The inverse distance weighting method, an estimation technique common to spatial data modeling, was used to estimate points in the fill grid of a grain cart prone to being immeasurable or highly variable. This method was successful in estimating missing points in a grain cart under difficult delivery conditions to within 15 cm of underestimation and 25 cm of overestimation. A model to predict the weight of grain in a grain cart was developed using the average grain height measured in the cart. The model demonstrated high robustness by producing mean errors that changed by less than 2% of the total cart volume when the delivery conditions strayed from typical conditions to highly biased conditions. The decision support system that was developed exhibited robust performance when critical features of the system were tested at typical levels. Field testing validated the potential to apply the decision support system to autonomous combine unloading systems by producing predictable and consistent final cart volumes that were within 5% of the total volume.