{"id":{"repo_id":"cau-kiel","oai_identifier":"oai:macau.uni-kiel.de:macau_mods_00008094"},"canonical_url":"https://search.dev.ndltd.org/etd/cau-kiel/oai:macau.uni-kiel.de:macau_mods_00008094","repository":{"repo_id":"cau-kiel","name":"Christian-Albrechts Universität Kiel","base_url":"https://macau.uni-kiel.de/servlets/OAIDataProvider"},"display":{"title":"From Passive Data Collection to Sensor-Level Intelligence","abstract":"The rapid proliferation of IoT has led to billions of connected devices operating across domains such as smart cities, agriculture, and environmental monitoring. Many of these devices are deployed in resource-constrained environments where stable power, high bandwidth, and continuous connectivity cannot be guaranteed. Traditional cloud-centric architectures, which rely on streaming raw sensor data to the cloud, impose high communication costs, accelerate battery depletion, and limit scalability. To overcome these limitations, this thesis integrates TinyML into resource-constrained IoT devices, enabling sensor-level intelligence that reduces reliance on continuous data transmission while extending device lifetime, conserving bandwidth, and preserving efficiency. By performing local analytics, devices can decide not only what information to transmit but also when transmission is necessary, forming event-triggered communication strategies that reduce overhead while maintaining accuracy and timeliness. The research advances two complementary directions. The first leverages compact TinyML models for on-device event detection with emphasis on seismic monitoring such as earthquake detection, ensuring that only high-value information is sent to the cloud. The second direction focuses on environmental monitoring, where maintaining a continuous system view in the cloud is essential. Instead of streaming data at the sensor sampling rate, an event-triggered approach is employed in which sensor nodes transmit only when observations deviate from expected trends. The cloud uses predictive models to estimate intermediate values until real data are received, minimizing communication without compromising accuracy. Together, these approaches establish a foundation for scalable, energy-efficient IoT systems that extend device lifetime, reduce communication bottlenecks, and enable reliable monitoring in environments where conventional cloud-centric architectures fail.","abstract_html":"The rapid proliferation of IoT has led to billions of connected devices operating across domains such as smart cities, agriculture, and environmental monitoring. Many of these devices are deployed in resource-constrained environments where stable power, high bandwidth, and continuous connectivity cannot be guaranteed. Traditional cloud-centric architectures, which rely on streaming raw sensor data to the cloud, impose high communication costs, accelerate battery depletion, and limit scalability. To overcome these limitations, this thesis integrates TinyML into resource-constrained IoT devices, enabling sensor-level intelligence that reduces reliance on continuous data transmission while extending device lifetime, conserving bandwidth, and preserving efficiency. By performing local analytics, devices can decide not only what information to transmit but also when transmission is necessary, forming event-triggered communication strategies that reduce overhead while maintaining accuracy and timeliness. The research advances two complementary directions. The first leverages compact TinyML models for on-device event detection with emphasis on seismic monitoring such as earthquake detection, ensuring that only high-value information is sent to the cloud. The second direction focuses on environmental monitoring, where maintaining a continuous system view in the cloud is essential. Instead of streaming data at the sensor sampling rate, an event-triggered approach is employed in which sensor nodes transmit only when observations deviate from expected trends. The cloud uses predictive models to estimate intermediate values until real data are received, minimizing communication without compromising accuracy. Together, these approaches establish a foundation for scalable, energy-efficient IoT systems that extend device lifetime, reduce communication bottlenecks, and enable reliable monitoring in environments where conventional cloud-centric architectures fail.","abstract_has_math":false,"creators":["Zainab, Tayyaba"],"institution":"Christian-Albrechts-Universität zu Kiel","degree_name":null,"degree_level":"thesis.doctoral","degree_discipline":null,"degree_department":null,"school":null,"contributors":["Landsiedel, Olaf","Förster, Anna"],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2026,"date_issued":"2026-03-05","date_published":"2026-03-05","updated_at":"2026-07-24T01:35:31Z","subjects":["Event-triggered communication","TinyML","Deep Neural Networks","Low-Power","Internet of Things","Edge AI","On-device","Seismological data analysis","Earthquake detection","Time-series forecasting"],"languages":[],"rights":[],"rights_urls":[],"identifier_entries":[]},"links":{"outbound_url":"https://macau.uni-kiel.de/receive/macau_mods_00008094","outbound_label":"Repository record","outbound_source":"source_url"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor","label":"Contributor","values":["Landsiedel, Olaf","Förster, Anna"]},{"key":"dc:creator","label":"Author","values":["Zainab, Tayyaba"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:publisher","label":"Institution","values":["Universitätsbibliothek Kiel"]},{"key":"dc:type","label":"Dc Type","values":["PhDThesis"]},{"key":"thesis:degree_level","label":"Degree Level","values":["thesis.doctoral"]},{"key":"thesis:institution_name","label":"Thesis Institution Name","values":["Christian-Albrechts-Universität zu Kiel"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Event-triggered communication","TinyML","Deep Neural Networks","Low-Power","Internet of Things","Edge AI","On-device","Seismological data analysis","Earthquake detection","Time-series forecasting"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["The rapid proliferation of IoT has led to billions of connected devices operating across domains such as smart cities, agriculture, and environmental monitoring. Many of these devices are deployed in resource-constrained environments where stable power, high bandwidth, and continuous connectivity cannot be guaranteed. Traditional cloud-centric architectures, which rely on streaming raw sensor data to the cloud, impose high communication costs, accelerate battery depletion, and limit scalability. To overcome these limitations, this thesis integrates TinyML into resource-constrained IoT devices, enabling sensor-level intelligence that reduces reliance on continuous data transmission while extending device lifetime, conserving bandwidth, and preserving efficiency. By performing local analytics, devices can decide not only what information to transmit but also when transmission is necessary, forming event-triggered communication strategies that reduce overhead while maintaining accuracy and timeliness. The research advances two complementary directions. The first leverages compact TinyML models for on-device event detection with emphasis on seismic monitoring such as earthquake detection, ensuring that only high-value information is sent to the cloud. The second direction focuses on environmental monitoring, where maintaining a continuous system view in the cloud is essential. Instead of streaming data at the sensor sampling rate, an event-triggered approach is employed in which sensor nodes transmit only when observations deviate from expected trends. The cloud uses predictive models to estimate intermediate values until real data are received, minimizing communication without compromising accuracy. Together, these approaches establish a foundation for scalable, energy-efficient IoT systems that extend device lifetime, reduce communication bottlenecks, and enable reliable monitoring in environments where conventional cloud-centric architectures fail."]},{"key":"dc:format.medium","label":"Dc Format Medium","values":["application/pdf"]},{"key":"dc:title","label":"Title","values":["From Passive Data Collection to Sensor-Level Intelligence"]}]}],"canonical_facts":{"dc:contributor":["Landsiedel, Olaf","Förster, Anna"],"dc:creator":["Zainab, Tayyaba"],"dc:description.abstract":["The rapid proliferation of IoT has led to billions of connected devices operating across domains such as smart cities, agriculture, and environmental monitoring. Many of these devices are deployed in resource-constrained environments where stable power, high bandwidth, and continuous connectivity cannot be guaranteed. Traditional cloud-centric architectures, which rely on streaming raw sensor data to the cloud, impose high communication costs, accelerate battery depletion, and limit scalability. To overcome these limitations, this thesis integrates TinyML into resource-constrained IoT devices, enabling sensor-level intelligence that reduces reliance on continuous data transmission while extending device lifetime, conserving bandwidth, and preserving efficiency. By performing local analytics, devices can decide not only what information to transmit but also when transmission is necessary, forming event-triggered communication strategies that reduce overhead while maintaining accuracy and timeliness. The research advances two complementary directions. The first leverages compact TinyML models for on-device event detection with emphasis on seismic monitoring such as earthquake detection, ensuring that only high-value information is sent to the cloud. The second direction focuses on environmental monitoring, where maintaining a continuous system view in the cloud is essential. Instead of streaming data at the sensor sampling rate, an event-triggered approach is employed in which sensor nodes transmit only when observations deviate from expected trends. The cloud uses predictive models to estimate intermediate values until real data are received, minimizing communication without compromising accuracy. Together, these approaches establish a foundation for scalable, energy-efficient IoT systems that extend device lifetime, reduce communication bottlenecks, and enable reliable monitoring in environments where conventional cloud-centric architectures fail."],"dc:format.medium":["application/pdf"],"dc:publisher":["Universitätsbibliothek Kiel"],"dc:subject":["Event-triggered communication","TinyML","Deep Neural Networks","Low-Power","Internet of Things","Edge AI","On-device","Seismological data analysis","Earthquake detection","Time-series forecasting"],"dc:title":["From Passive Data Collection to Sensor-Level Intelligence"],"dc:type":["PhDThesis"],"thesis:degree_level":["thesis.doctoral"],"thesis:institution_name":["Christian-Albrechts-Universität zu Kiel"]},"updated_at":"2026-07-24T01:35:31Z"}