Vegetables and other row-crops represent a large share of the agricultural production. There is a large variation in crop species, and a limited availability in specialized herbicides. The robot presented here utilizes the systematic cultivation techniques of row crops to navigate and operate in the eld. By the use of machine vision it separates seeded vegetable crops from weed, and treat each weed within the row with individual herbicide droplets, without aecting the crop. This Dropon-Demand (DoD) method allow the use of non-selective herbicides and signicant reductions in herbicide use.This thesis presents six research papers concerning the development of the DoD system and the mobile robot. The robot is tailored to its purpose with cost, maintainability, ecient operation and robustness in mind. The three-wheeled design is unconventional, and the design maintains maneuverability and stability with the benet of reduced weight, complexity and cost. The DoD system for herbicide application has been developed within and in connection with this project. The inuence of liquid properties viscosity and surface tension on the formation and stability of droplets has been tested in lab trials. A control circuit for synchronized control of solenoid valves was developed and tested.Indoor pot trials with four weed species demonstrated that the Drop-on-Demand system (DoD) could control the weeds with as little as 7.6 µg glyphosate or 0.15 µg iodosulfuron per plant. The results also highlight the importance of liquid characteristics for leaf retention, as the common herbicide glyphosate had no eect unless mixed with suitable additives. The trials document the DoD eect on weed species not previously described in literature, and with an alternative herbicide to glyphosate, iodosulfuron. A eld trial with the robot was performed in a carrot eld, and all the weeds were eectively controlled with the DoD system.iii Abstract The robot and DoD system represent an important contribution to the range of systems presented witin Precision Agriculture for in-row weed control -a movement which as a whole represent a paradigm shift to the environmental impact and health risks of weed control, while providing valuable new tools to the producers. My advisor, Prof. Jan Tommy Gravdahl has the ability to delve deeply into a wide array of technical elds. He has been a solid and consistent support, which has been vital for me to bridge the divide between the industrial and academic worlds. I will strive for the opportunities to work more together.Friendships have grown both at NTNU, UMN and Adigo which will outlast this thesis. I thank you, my family, colleagues and friends for the support to complete this journey. And not the least, I thank my wife Camilla. This introduction will describe the background for the project and its motivation, and the context and relationship between each paper is described in Section 1.4. The conclusion in Chapter 3 will tie together our ndings and describe our perspective on the path forwards for Drop on De...
This work addresses the implementation issue of constrained Model Predictive Control (MPC) for the autonomous trajectory-tracking problem. The chosen process to control is a Wheeled Mobile Robot (WMR) described by a discrete, Multiple Input Multiple Output (MIMO), state-space and linear parameter varying kinematic model. The main motivation of the constrained MPC usage in this case relies on its ability in considering, in a straightforward way, control and states constraints that naturally arise in trajectory tracking practical problems. The efficiency of the presented control scheme is validated through experimental results on a two wheeled mobile robot using both STM32F429II and STM32F407ZG microcontrollers. The controller implementation is facilitated by the usage of the automatic C code generation and interesting optimization before real-time execution. Based on the experimental results obtained, the good performance and robustness of the proposed control scheme are established.
Abstract-Drop-on-demand weed control is a field of research within Precision Agriculture, where the herbicide application is controlled down to individual droplets. This paper focuses on the fluid dynamics and electronics design of the droplet dispensing. The droplets are formed through an array of nozzles, controlled by two-way solenoid valves.A much used control circuit for opening and closing a solenoid valve is a spike and hold circuit, where the solenoid current finally is discharged over a Schottky diode on closing. This paper presents a PWM design, where the discharge is done by reversing the polarity of the voltage. This demands an accurate timing of the reverse spike not to recharge and reopen the valve. The PWM design gives flexibility in choosing the spike and hold voltage arbitrarily, and may use fewer components. Calculations combined with laboratory experiments verify this valve control strategy.In early flight the stability of the tail, or filament, is described in theory by the Ohnesorge number. In later flight, when a droplet shape has formed, the droplet stability is governed by the Weber number. These two considerations have opposite implications on the desired surface tension of the fluid. The Weber number is more important for longer distances, as the filament satelites normally catch up and join the main droplet in flight.
The field of precision agriculture increasingly utilize and develop robotics for various applications, many of which are dependent on high accuracy localization and attitude estimation. Special attention has been put towards full attitude estimation by low-cost sensors, in relation to the development of an autonomous field robot. Quaternions have been chosen due to its continuous nature, and with respect to applications in the pipeline with on other platforms. The performance and complexity of two approaches to attitude estimation has been investigated: One Multiplicative Extended Kalman Filter (MEKF) and one non-linear observer. Both were implemented on an ARM Cortex M3 microcontroller with sensors for a Attitude Heading Reference System (AHRS), and benchmarked towards a relative high grade commercial AHRS device. The relative computational burden of the MEKF have been underlined, by execution times more than 10 times those of the non-linear estimator. The implementation complexity is also significantly lower for the non-linear observer, which facilitate test and verification through more transparent software.
Visual Odometry (VO) is increasingly a useful tool for robotic navigation in a variety of applications, including weed removal for agricultural robotics. The methods of evaluating VO are often computationally expensive and can cause the VO measurements to be significantly delayed with respect to a compass, wheel odometry, and GPS measurements. In this paper we present a Bayesian formulation of fusing delayed displacement measurements. We implement solutions to this problem based on the unscented Kalman filter (UKF), leading to what we term an unscented multi-point smoother. The proposed methods are tested in simulations of an agricultural robot. The simulations show improvements in the localization RMS error when including the VO measurements with a variety of latencies.
In the heavy asset industry, such as oil & gas, offshore personnel need to locate various equipment on the installation on a daily basis for inspection and maintenance purposes. However, locating equipment in such GPS denied environments is very time consuming due to the complexity of the environment and the large amount of equipment. To address this challenge we investigate an alternative approach to study the navigation problem based on visual imagery data instead of current ad-hoc methods where engineering drawings or large CAD models are used to find equipment. In particular, this paper investigates the combination of deep learning and decomposition for the image retrieval problem which is central for visual navigation. A convolutional neural network is first used to extract relevant features from the image database. The database is then decomposed into clusters of visually similar images, where several algorithms have been explored in order to make the clusters as independent as possible. The Bag-of-Words (BoW) approach is then applied on each cluster to build a vocabulary forest. During the searching process the vocabulary forest is exploited to find the most relevant images to the query image. To validate the usefulness of the proposed framework, intensive experiments have been carried out using both standard datasets and images from industrial environments. We show that the suggested approach outperforms the BoW-based image retrieval solutions, both in terms of computing time and accuracy. We also show the applicability of this approach on real industrial scenarios by applying the model on imagery data from offshore oil platforms.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.