136879 research outputs found
Sort by
Characterizing Biological Processes Influencing Inter-Animal Variation in Efficiency of Nutrient Utilization in Beef and Dairy Cattle
No full textResidual feed intake (RFI), a metric of feed efficiency, is moderately heritable and minimally associated with body size and productivity, making it an ideal trait for investigation. The first objective focused on characterizing the biological mechanisms associated with feed efficiency in cattle with divergent genomically enhanced breeding values for RFI (RFIg). Holstein heifers with low- (n=29) or high- (n=26) RFIg were utilized in an 84-d feeding trial. Low RFIg heifers consumed 8% less feed, demonstrated a more favorable RFI phenotype, and exhibited favorable feeding behavior compared to high RFIg heifers. Heifers with low RFIg produced 8% less methane and had 6% lower heat production than high RFIg heifers. Gas yields and digestibility did not differ between heifers with divergent RFIg. Low RFIg heifers exhibited greater rumen volatile fatty acid concentrations and had less enrichment of ruminal bacteria than high RFIg heifers; however, there were no differences in bacterial abundance between heifers with divergent RFIg. Amplicon sequence variants from the Prevotellaceae and Ruminococcaceae families differed between heifers with divergent RFIg. Variation in RFIg explained by feeding behavior, gas flux, digestibility and rumen fermentation, and complete blood count accounted for 56.3% of the variation in RFIg. Results suggest that implementation of RFIg may provide producers opportunities to select more economically and environmentally sustainable cattle. The second objective focused on developing an algorithm to assess feed bunk replacement events using the GrowSafe system. Utilizing growing beef steers (n=20), an electronic algorithm was developed to quantify feed bunk replacement events, with an optimal replacement criterion determined to be ���18 to ���21 s. Using the replacement algorithm, the effect of interactive and competitive feeding activity on performance, feed efficiency and feeding behavior was investigated. Data from growing beef steers (n=497; 3 trials) were used. A replacement activity index (RAI) was developed, and steers were classified into divergent phenotypes. Solitary (low RAI) steers had lower DMI, improved feed efficiency, and displayed differences in feeding behavior and diurnal feeding patterns compared to interactive (high RAI) steers. There is a need to further explore inherent sources of variation in biological processes and identify candidate biomarkers associated with RFI
Indoor Occupancy Detection and Tracking Using Networked Sensor Nodes
Get PDFIndoor occupancy detection and tracking are essential elements of occupant behavior research. One of the recent advancements in this area is the use of networked sensor nodes to create a more comprehensive occupancy picture of an indoor space where multiple sensor nodes can identify human presence while delivering superior accuracy compared to a system that relies on standalone sensor nodes. Standalone sensor nodes often produce false negative and positive detections, due to sensor coverage gaps and shortcomings in the underlying occupancy sensing technologies.
This research aims to improve indoor occupancy detection and tracking performance using network-level sensor fusion and occupancy estimation techniques. These techniques exploit the redundant, correlated, and complementary information in the time-series data that networked occupancy sensor nodes produce. Using a combination of node deployment configurations and indoor environments, several occupancy detection and tracking methods are proposed which can potentially contribute towards applications areas like Occupant comfort, Indoor health, Energy and space utilization, Building Design, and Occupant safety & security.
The presented work focuses on optimizing networked nodes-based occupancy detection and tracking pipeline. Some notable elements of a typical pipeline include data collection and labeling strategies, sensor models, data fusion techniques, node-level and network-level machine learning algorithms, and estimation algorithms. Each sensor node can use one or multiple node-level occupancy sensors technologies such as microphones, ambient temperature, humidity, carbon dioxide (CO2), and passive infrared (PIR) sensors.
The main contributions of this dissertation include: (i) A node-level on-device lifelong (ODLL) classifier is proposed that continuously learns evolving occupancy patterns over time; (ii) A network-level algorithm that exploits the inter-node spatial adjacency information and as well as observation correlations between the nodes; (iii) A 1-minute resolution occupancy tracking system that exploits the node adjacency and node correlation constraints to filter-out the unreachable occupancy states
