Ruminants play an essential role in supplying nutrients to the global population. Despite notable advancements in the livestock industry, there is a rising demand for animal protein products and a pressing need for sustainable practices. Consequently, it is imperative to focus on improving efficiency and sustainability across the environmental, economic, and social dimensions of the livestock system. Precision livestock farming (PLF) technologies have emerged as a potential solution to enhance sustainability by integrating individual animal monitoring and automated control over animal productivity, environmental impacts, health, and welfare parameters. Although PLF holds promise for improving livestock management practices, its widespread adoption is hindered by challenges including the high costs associated with implementation, data ownership, and implementation across different environments. he overarching aim of this research was to investigate and propose solutions to the challenges that limit the extensive implementation of wearable technologies in livestock systems. The primary objective of the first study was to develop and assess the utility of an open-source, low-cost research wearable technology equipped with Bluetooth for monitoring ruminants in a confined setting. The study successfully demonstrated the functionality and cost-effectiveness of this technology and its potential for monitoring ruminants' behavior in research and practical applications. Building upon the success of the technology in intensive systems, the subsequent study focused on updating the wearable sensor for deployment in extensive systems. This was achieved by incorporating LoRa data transmission and enabling real-time monitoring of livestock location. The study effectively demonstrated the feasibility of the updated technology for real-time monitoring of livestock in extensive grazing systems. In continuation of testing the feasibility of sensors, the subsequent experiment aimed to assess the accuracy and precision of a low-cost wearable sensor photoplethysmography (PPG) sensor in monitoring heart-rate (HR) of sheep housed under high-temperature conditions. The results revealed poor accuracy and precision in detecting HR changes using the PPG sensor. Future studies should explore alternative sensor deployment methods and data analytics techniques to improve the accuracy of a PPG sensor in detecting HR in livestock animals. The follow-up study focuses on evaluating the suitability of a continuous glucose monitor (CGM) designed for humans in measuring interstitial glucose concentrations in sheep, as a potential replacement for traditional blood glucose measurements. The findings demonstrated great potential of CGM in detecting changes in glucose concentrations in sheep. However, the study`s limitations such as the small sample size, warranting further investigation with a larger sample size and potential standardization with laboratory analysis bore implementing CGMs as a replacement for traditional glucose measurement methods in research. The limited expansion of technology application in extensive livestock systems, in contrast to confined operations, can be attributed to challenges such as limited battery life and data transmission. To overcome these limitations, edge processing techniques which involve performing data processing, analytics, and decision-making closer to the data source, have been proposed as cost-effective strategy for enhancing the usability of inertial measurement unit systems (IMU) in monitoring grazing animal behavior. Therefore, the objective of the fifth study was to explore different classification techniques suitable for edge processing using an open-source IMU. Analysis of variances, logistic regression, support vector machine, and random forest were evaluated for classifying grazing, walking, standing, and lying behaviors. The random forest model achieved the highest accuracy (93%) in classifying grazing using 1-minute interval. Moreover, the algorithms were compared considering a periodic snapshot of data with intervals of 3 or 5 seconds, and interesting revealed no significant impact on algorithm accuracy on differentiating behavior of grazing cows using IMU systems. Heat stress has negative impacts on animal behavior, welfare, and productivity. While IMU systems have been used to detect behavioral changes in thermoneutral conditions, their effectiveness on heat-stressed animals remains unclear. The objective of the last study was to investigate changes in sheep behavior using a low-cost IMU and the influence of ambient temperature in the algorithms ability to classify behaviors. Eating, lying, standing and ruminating while standing and lying were classified during exposure to different ambient temperature patterns. The algorithm demonstrated acceptable accuracies in differentiating behaviors under thermoneutral conditions, but its performance was impaired when tested outside the thermal range. Future research should focus on developing algorithms that account for different environmental conditions to improve the accuracy of IMU in classifying animal behavior. Collectively, these investigations contribute to enhancing the applicability of technologies in livestock systems. / Doctor of Philosophy / The global population relies on ruminant animals, such as cattle and sheep, for essential nutrient. However, with the increasing demand for animal protein products, there is a growing need for sustainable practices in the livestock industry. Precision livestock farming (PLF) technologies have emerged as a potential solution to enhance sustainability by enabling individual animal monitoring. However, challenges such as data ownership and accessibility and high costs, impair its adoption. To overcome these challenges and enhance the applicability of wearable sensors in livestock systems, this research aimed to explore potential solutions. The objective of the first study was to develop and evaluate an open-source, low-cost wearable technology equipped with Bluetooth for monitoring ruminants in confined settings. The study successfully demonstrated the functionality and cost-effectiveness of this technology for monitoring ruminant behavior. Building up the success of the technology in intensive systems, the subsequent study focused on updating the wearable sensor for deployment in extensive systems. This was achieved by incorporating LoRa data transmission, enabling real-time monitoring of livestock. The study effectively demonstrated the feasibility of and potential of the updated technology for real-time monitoring in extensive livestock systems. Continuing with the feasibility testing of technologies, the next experiment aimed to assess the accuracy and precision of a low-cost photoplethysmography (PPG) sensor in monitoring heart rate (HR) in sheep housed under high-temperature conditions. Unfortunately, the results indicated poor accuracy and precision in detecting HR changes using the PPG sensor. Future studies should explore alternative sensor deployment methods and data analysis techniques to improve the accuracy of PPG sensors for HR monitoring in livestock animals. The followed study focused on evaluating the suitability of a continuous glucose monitors (CGM) designed for humans to measure interstitial glucose concentrations in sheep and potentially replacing traditional blood glucose measurements. The findings demonstrated the potential of CGMs to detect changes in glucose but limitations such as the small sample size suggest the need for further investigations with a larger sample size and potential standardization with laboratory analysis before implementing CGM as a replacement for traditional glucose measurement methods in research. In extensive systems, where technology adoption has been slower compared to confined operations, edge processing techniques are proposed as a cost-effective strategy to monitor grazing animal behavior using inertial measurement unit systems (IMU). In the fifth study, different classification techniques were explored using an open-source IMU, including analysis of variances, logistic regression, support vector machine, and random forest. The random forest model achieved high accuracy (93%) in classifying grazing behavior with a 1-minute interval. Surprisingly, algorithm accuracy was not affected when snapshot in time was performed. The final study focused on using a low-cost IMU to investigate sheep behavior under varying ambient temperature conditions. While algorithm performed well under thermoneutral conditions, its accuracy decreased outside the thermal range. Future research should focus on algorithms that account for different environmental conditions to improve IMU accuracy in classifying behavior. These investigations contribute to enhancing technology's applicability to in livestock systems by addressing challenges and developing practical solutions to improve livestock management and animal well-being.

Crayfish provide a model which is simple, has an easily-accessible cardiovascular system and can be maintained in the laboratory conditions; the model has good utility for water quality assessment and ethophysiological studies. A noninvasive crayfish cardiac and behavioral activities monitoring (NICCBAM) system is discussed in the thesis. The system is inexpensive, has relatively few components and permits long-term continuous simultaneous monitoring of cardiac and behavioral activities of several crayfish. Moreover, compared to other available systems, it provides a novel approach of cardiac activity shape analysis which allows improving monitoring accuracy as well as obtaining additional information on crayfish functional state. The NICCBAM system was evaluated by comparing with the well-known electrocardiography system which demonstrated that cardiac contractions with both approaches were synchronous and that both signal shapes were similar. Experiments on crayfish cardiac activity relative to selected odors and chemicals demonstrated the promising potential of cardiac signal shape analysis, not only for detecting changes in the aquatic environment, but also for their classification.

Master of Science / Department of Clinical Sciences / Robert L. Larson / Bradley J. White / We performed research that investigated the ability of three dimensional accelerometers to classify cattle behavior and also describe the circadian patterns within that behavior. The first of three studies (validation study) tested a decision tree classification system and its ability to describe behaviors of lying, standing, and walking. Classification accuracies for lying, standing, and walking behaviors were 99.2%, 98.0%, and 67.8% respectively, with walking behavior having significantly lower accuracy (P<0.01). This study also tested the accuracy of classifying the above behaviors using different device reporting intervals, or epochs. Reporting intervals of 3, 5, and 10 seconds (s) were evaluated in their ability to describe cattle behaviors of lying, standing, and walking. Classification accuracies for the 3s, 5s, and 10s reporting interval were 98.1%, 97.7%, and 85.4% respectively, with no difference in classification accuracy of the 3s and 5s epochs (P=0.73) while the 10s epoch exhibited significantly lower overall accuracy (P<0.01). This validated accelerometer monitoring system was then implemented in two studies (Winter 2007 and Spring 2008) where the devices were used to describe behavior patterns of beef calves in a drylot production setting. Lying behavior of the cattle was analyzed and found to be significantly associated (P<0.001) with hour of the day. Calves in these studies spent most (> 55%) of the nighttime hours (2000 to 0400) involved in lying behavior and spent the least percentage of time lying (<30%) during periods of time where feed was presented at the bunk (0700 and 1700). Mean lying time was also associated with trial day (P<0.01) and most trial days (67.5%) calves spending between 45% and 55% of time lying. Variation of lying behavior was found between individuals (range 29% to 66%); however, consistency in lying behavior was found within individual calves across study periods. The accelerometer monitoring system studies presented here provide evidence these devices have utility in recording behaviors (lying, standing, and walking) of individual beef calves raised in typical production settings.

Schools are facing an increasing pressure to deal effectively with students' problem behaviors in the school environment. Research suggests that Behavior Monitoring Programs (BMPs) are effective and efficient secondary interventions to use in remedying problem behavior in the classroom and are acceptable to teachers, parents, and students. Most of the research on BMPs has been conducted at the elementary school level. The current study investigated the effectiveness of a BMP within a school-wide system of Positive Behavior Support (PBS) with three suburban high school students. Problem behaviors for each student were targeted based upon previous office discipline referral data (ODR) and teacher comments, and three behavioral goals were made for students based upon these findings, along with teacher input. Effectiveness of the intervention was measured by the increase in teacher's behavioral ratings on the Daily Behavior Report Card (DBRC). Furthermore, teachers, parents and students rated the intervention's effectiveness via a five-item intervention acceptability questionnaire. Results of the study suggest that the BMP intervention is both effective and acceptable for use with secondary students. All students experienced an increase in behavioral ratings on the DBRC during intervention. Across all students and all behaviors, the intervention resulted in an overall mean improvement of 63% in problem behaviors in the classroom. Average effect sizes were large while probability levels were low. Furthermore, all teachers, parents, and students rated the intervention as being acceptable. The average rating that all parents gave for all five items (on a 6 point scale with higher numbers indicating greater acceptability) was 5.2, while the average for students was 4.3. The student's teachers together rated all five items as 4.8.

In Multi-Agenten-Systemen (MAS) kooperieren und konkurrieren Agenten um ihre jeweiligen Ziele zu erreichen. Für optimierte Agenten-Interaktionen sind Kenntnisse über die aktuellen und zukünftigen Handlungen anderer Agenten (Interaktionsparter, IP) hilfreich. Bei der Ermittlung und Nutzung solcher Kenntnisse kommt dem automatischen Erkennen und Verstehen sowie der Vorhersage von Verhalten der IP auf Basis von Beobachtungen besondere Bedeutung zu. Die Dissertation beschäftigt sich mit der automatischen Bestimmung und Vorhersage von Verhalten der IP durch einen Modellierenden Agenten (MA). Der MA generiert fallbasierte, adaptive Verhaltens-Modelle seiner IP und verwendet diese zur Vorhersage ihrer Verhalten. Als Anwendungsszenario wird mit dem virtuellen Fußballspiel des RoboCup ein komplexes und populäres MAS betrachtet. Der Hauptbeitrag dieser Arbeit besteht in der Ausarbeitung, Realisierung und Evaluierung eines Ansatzes zur automatischen Verhaltens-Modellierung für ein komplexes Multi-Agenten-System. / In multi-agent-systems agents cooperate and compete to reach their personal goals. For optimized agent interactions it is helpful for an agent to have knowledge about the current and future behavior of other agents. Ideally the recognition and prediction of behavior should be done automatically. This work addresses a way of automatically classifying and an attempt at predicting the behavior of a team of agents, based on external observation only. A set of conditions is used to distinguish behaviors and to partition the resulting behavior space. From observed behavior, team specific behavior models are then generated using Case Based Reasoning. These models, which are derived from a number of virtual soccer games (RoboCup), are used to predict the behavior of a team during a new game. The main contribution of this work is the design, realization and evaluation of an automatic behavior modeling approach for complex multi-agent systems.