COLLABORATIVE RESEARCH: CPS: MEDIUM: EBEEVET: AN ELECTRONIC BEE-VETERINARIAN FRAMEWORK TO SECURE HONEY BEES

• Sponsoring Institution: National Institute of Food and Agriculture
• Project Status: ACTIVE
• Funding Source: AFRI COMPETITIVE GRANT
• Reporting Frequency: Annual
• Accession No.: 1033080
• Grant No.: 2024-67021-43696
• Cumulative Award Amt.: $300,000.00
• Proposal No.: 2024-07277
• Project Start Date: Sep 15, 2024
• Project End Date: Sep 14, 2027
• Grant Year: 2024
• Program Code: [A7302]- Cyber-Physical Systems

Recipient Organization
Carnegie Mellon University
5000 Forbes Avenue
Pittsburgh,PA 15213-3815

Performing Department
(N/A)

Non Technical Summary
Honey bees are not only of crucial importance for ecosystem stability but also for the human food supply. Their services are required for the production of more than 80 crops of agricultural interest or about a third of what we eat. Their yearly global value has been estimated to be up to $550 billion globally and $29 billion for the United States. However, drastic declines of honeybee populations have been documented over thepast two decades.This project proposes to develop and test Electronic Bee-Veterinarian (EbeeVet), a functional sensorframework that is specifically designed for bee hives and is fully integrated into the hive setups used by the majority of beekeepers, known as Langstroth hives. The framework will consist of machine learning (ML) methods to analyze a wealth of data and to propose timely and appropriate solutions, as well as scalable data management systems for multiple hives, and apiaries and cost-effective and self-sustainable sensing devices.

Animal Health Component

30%

Research Effort Categories

Basic

40%

Applied

30%

Developmental

30%

Classification

Knowledge Area (KA)	Subject of Investigation (SOI)	Field of Science (FOS)	Percent
307	3010	2080	100%

Knowledge Area
307 - Animal Management Systems;

Subject Of Investigation
3010 - Honey bees;

Field Of Science
2080 - Mathematics and computer sciences;

Keywords

agriculture

data analytics and machine learning

Goals / Objectives
How can machine learning technology be used to maximize the health of honey bee colonies?Honey bees are not only of crucial importance for ecosystem stability but also for the human food supply.Their services are required for the production of more than 80 crops of agricultural interest or about a thirdof what we eat. Their yearly global value has been estimated to be up to $550 billion globally and $29 billionfor the United States. However, drastic declines of honeybee populations have been documented over thepast two decades.This project proposes to develop and test Electronic Bee-Veterinarian (EbeeVet), a functional sensorframework that is specifically designed for bee hives and is fully integrated into the hive setups used by themajority of beekeepers, known as Langstroth hives. The framework will consist of machine learning (ML)methods to analyze a wealth of data and to propose timely and appropriate solutions, as well as scalabledata management systems for multiple hives, and apiaries and cost-effective and self-sustainable sensingdevices. The PI team consists of a unique combination of experts, including biologists, computer engineers,and machine learning experts. The team has collected over 10 years' worth of sensor data (temperature,humidity, etc) from multiple bee colonies, providing a solid foundation for in-depth analysis and significanttechnological advancements.

Project Methods
** EFFORTSWe plan to develop state-of-the-art machine-learning methods for measuring and monitoring the health of bee-hives. The goal is to issue early-warnings to the bee-keepers, when we notice anomalous behavior, like unexpected temperature or humidity inside the hive. As mentioned in the proposal, we will continue our practice of disseminating our results through graduate courses, bee-health conferences,and publications in machine learning and database conferences and journals.** EVALUATIONOur project involves a thorough plan for individual thrust evaluation as well as integrated system validation. We will implement our approaches and provide tools, documents, and datasets in public.* Testbeds.We will leverage the experimental apiary located at UCR for our testbed. This is being managed by the CIBER center led by PI Baer and other PIs Kim and Tsotras also have access to the apiary and have been working with beekeepers and entomologists to help understand honeybee health using wireless sensors and real-time data analysis and control.* Evaluation of Individual Tasks (Thrusts 1 - 3).Each of the research tasks will first be evaluated individually relative to existing results in the relevant literature. We plan to make use of open-source tools, datasets, and the aforementioned testbed in the individual evaluation.* Integrated System Evaluation (Thrust 4).We will follow the procedures discussed in Sec. 4 for system integration and validation. The results from the UCR testbed will be shared with our beekeeper collaborators (see Facilities and letters of collaboration).

Progress 09/15/24 to 09/14/25

Outputs
Target Audience:The target audience is cross-disciplinary: computer scientists (machine learning and database); electrical engineers (sensor design and control theory); entomologists (phenotypic biomarkers for bee health);and beekeepers. Changes/Problems: Nothing Reported What opportunities for training and professional development has the project provided?The project has provided training and professional development to the following PhD students: Mst Shamima Hossain, Sakshi Watts, Genesis Chong at UCR, Meng-Chieh (Jeremy) Lee and Karish Grover at CMU. We had weekly research meetings where all project members from UCR and CMU participated. The students gained expertise in analyzing beehive time-series sensor data and designing novel methods for proactive hive microclimate control to maximize honey yield and colony health. The PhD student involved in this project (Mst Shamima Hossain) presented our contributions at several research competitions and earned numerous accolades: UCR Grad Slam 2025 Finalist (Honorable Mention) Best Presentation Award at ACM/IEEE CPS-IoT Week Ph.D. Forum 2025 Best Presentation Award at UCR ORCA Forum 2025 How have the results been disseminated to communities of interest?The research findings have been disseminated through publications in premier journals: Mst Shamima Hossain, Christos Faloutsos, Boris Baer, Hyoseung Kim, and Vassilis J. Tsotras. 2025. Principled Mining, Forecasting, and Monitoring of Honeybee Time Series with EBV+. ACM Trans. Knowl. Discov. Data 19, 5, Article 100 (June 2025), 30 pages. https://doi.org/10.1145/3719014 We have open-sourced the implementation and dataset for our project: https://github.com/rtenlab/EBeeVet Our EBV+ framework has been featured in over 20 local and national media outlets: Following massive colony loss in early 2025, new methods analyzing temperature data help beekeepers predict issues in the hive.The California Aggie, May 2025. (link) Superbugs, Indigenous video games, and tipping dilemmas.UCR Press Release, April 2025. (link) Harnessing Computer Science to Protect Bee Populations.Bioengineer.org, March 2025. (link) Using Computer Science To Save the Bees Researchers: Use Sensors, Forecasting Models To Track Honeybee Health.CMU Press Release, March 2025. (link) Using Computer Science To Save the Bees.ENN: Environmental News Network, March 2025. (link) Using computer science to save the bees.ScienceDaily, March 2025. (link) Computer science majors 'buzzy' with saving the bees.The Appalachian, March 2025. (link) How Computer Science Is Using Sensors And Forecasting Models To Save Bees And Protect Pollination.Quantum Zeitgeist, March 2025. (link) New technology could save declining honeybee populations.Tech Explorist, March 2025. (link) New AI-Powered Sensors May Hold the Key to Saving Honeybees.MSN, March 2025. (link) Beehive sensors offer hope in saving honeybee colonies: Innovation allows for remote monitoring of beehive health.UCR Press Release, February 2025. (link) Electronic Bee-Veterinarian: UC Riverside Develops Beehive Sensors to Save Colonies.Eureka! UK, February 2025. (link) Beehive sensors offer hope in saving honeybee colonies: Innovation allows for remote monitoring of beehive health.EurekaAlert!, February 2025. (link) Beehive Sensors: A Promising Solution for Saving Honeybee Populations.Bioengineer.org, February 2025. (link) Beehive sensors could save rapidly declining honeybee populations.DPA Magazine, February 2025. (link) Beehive sensors offer hope in saving honeybee colonies.Phys.org, February 2025. (link) Electronic Bee-Veterinarian helps maintain healthy hives.TheEngineer, February 2025. (link) Beehive sensors offer hope in saving honeybee colonies.ENN: Environmental News Network, February 2025. (link) New Low-Cost Beehive Sensors Could Help Save Honeybee Colonies.EcoWatch, February 2025. (link) Beehive sensors offer hope in saving honeybee colonies.ScienceDaily, February 2025. (link) Beehive sensors offer hope in saving honeybee colonies.GroundNews, February 2025. (link) What do you plan to do during the next reporting period to accomplish the goals?Building on our successful EBV+ monitoring framework, we plan to develop an automatic end-to-end hive thermal regulation framework that integrates heating/cooling actuators with minimal in-hive sensing. This is important because temperature control improves honey yield: beekeepers often assist bees' thermoregulation using supplemental feeding (e.g., sugar water) and installing electronic heating pads for cold. Previous studies have shown external temperature assistance directly boosts honey production as well. We will work on techniques to predict hive core temperatures using in-hive sensors and easily-obtainable meteorological data, to estimate honey production based on bee thermoregulation energy costs, and to control heating/cooling actuators to maximize honey yield. We will also construct hardware setups to deploy our automatic control methods in the field. We will continue our collaborative approach with weekly meetings between UCR and CMU teams to ensure integrated and coherent research outcomes.

Impacts
What was accomplished under these goals? During this reporting period, we made significant progress across multiple components of the EbeeVet framework: We developed the Electronic Bee-Veterinarian Plus (EBV+) method, a novel approach for analyzing hive health based on thermal diffusion equations and a sigmoid feedback-loop controller. This method addresses key limitations of existing approaches by providing: (i) effective analysis of multiple real-world beehive time series data (both recorded and streaming), (ii) explainable results with interpretable parameters such as "hive health factor" that beekeepers can easily understand and trust, (iii) proactive alert capabilities that warn beekeepers before homeostasis disruptions become detrimental to colony health, and (iv) computational scalability for large-scale deployment. Experimental evaluation on real-world data showed substantial improvements over existing methods (ARX, seasonal ARX, Holt-winters, and DeepAR), with up to 72% higher forecasting accuracy while using 600 times fewer parameters. Our method successfully detected anomalies and generated alerts that align with domain expert assessments. The computational efficiency of our method allowed reconstruction of two months of sensor data in less than 1 minute on standard laptop hardware. These advances were published in ACM Transactions on Knowledge Discovery from Data (TKDD) (Hossain et al. 2025), which is a top-tier venue for data mining research. We also filed a provisional patent application in April 2025 (No. 63/790617).

Publications

• Type: Peer Reviewed Journal Articles Status: Published Year Published: 2025 Citation: Shamima Hossain, Christos Faloutsos, Boris Baer, Hyoseung Kim, Vassilis J. Tsotras: Principled Mining, Forecasting, and Monitoring of Honeybee Time Series with EBV+. ACM Trans. Knowl. Discov. Data 19(5): 1-30 (2025)