Robotic farming on marginal, highly sloped lands

Abstract

One of the most pressing issues of our time is how to feed around 9.7 billion people by 2050. Cropland expansion is one of the leading factors in global agricultural production growth to meet the rising demands of an escalating population. Arable steep grassland, hills or uneven terrain present difficulties to farming with large conventional agriculture machinery and equipment’s. The current technology is unsafe and unsuitable to operate on sloping terrain. This technological barrier to slope farming has prevented thousands of hectares of arable land from being cultivated, primarily in the United States Great Plains region. Therefore, we proposed a fleet of small Autonomous Ground Vehicles (AGVs) to expand farming to marginal, uneven, and highly sloped terrain. The proposed fleet aims to perform the essential agricultural operations ranging from seeding to harvesting on sloping terrain. The research aimed to explore the potential, capabilities, and limitations of small ground vehicles to perform the sloped crop work. The dissertation consisted of five chapters. The first chapter introduced the undertaken problem, background, and proposed solution. It also outlined the included chapters with goals and significance. The second chapter laid the foundation of a multi-AGV fleet by determining the single AGV’s suitability and capabilities and by quantifying its physical limits for sloped crop work in a controlled soil bin setup. A standard drawbar pull test was performed in a soil bin to evaluate the AGV’s performance against the varying slope, speed, and drawbar. The AGV delivered optimum power efficiency and generated enough drawbar pull with optimum travel reduction. The results found that the prototype AGV can successfully operate on slopes up to 18º, indicating that high-sloped terrain or hills could be farmed with the proposed system. The vehicle behavior models in a sloping environment are essential for fleet operation, path planning, and developing a control algorithm. Hence, Chapter 3 aimed to develop the AGV’s behavior models from laboratory soil bin data. Artificial neural network (ANN) models were developed. Shallow ANNs were fast, accurate, and reliable tools to predict AGV behavior in a controlled laboratory setup (i.e., sloped soil bin). The predictive AGV’s behavior model from a control laboratory setup proved to be an excellent starting point for optimizing the vehicle control parameters. However, these models cannot be extended to predict the AGV’s behavior in a continuously varying slope environment. Therefore, Chapter 4 aimed to develop machine learning-based models on data collected from a real-field environment. Machine learning and deep learning-based models were developed and analyzed. The study found that the deep neural networks (DNN) model was well-suited for predicting the AGV’s behavior in a sloped, real-field environment. Chapters 3 and 4 explored the capabilities of an artificial intelligence methods to simulate the AGV’s behavior on sloping terrain. The developed models predicted the AGV’s specific dynamic response, including traction, slip, and energy from the inputs of AGV’s velocity, applied load, and slope. A small and lightweight AGV was unable to provide the downforce and drawbar required for a traditional seeder. Hence, these AGVs would need a specialized robotic grain drill. The feed mechanism is the heart of the grain drill, and its design and performance influence the plant population and crop yield. Therefore, Chapter 5 aimed to design and develop a screw auger type feed mechanism. The feed mechanism was developed and tested in a laboratory setup against speed, vibration, and slope as control variables. The study delivered a bulk feed mechanism for wheat drilling, which can be easily scaled and adopted by small autonomous vehicles or mobile robots. The dissertation laid the foundation for robotic farming on the sloped terrain, and the envisioned multi-AGV fleet may provide a valid solution to farm the arable uneven, highly sloped terrain. The findings provide a groundwork for robotics and automation, which has the potential to solve the emerging problems in the food production system by producing food, fuel, and fiber for the growing population.