Project Background: IoT is crucial for smart farming where IoT sensors can gather real-time data on soil, weather, livestock, and crops enabling real-time monitoring and management of agriculture products and resources. Due to long-term cost, sustainability, and environmental concerns, the energy harvesting technology has become popular battery substitute to power IoT for sensing and intelligent processing in a sustainable way. The sustainable IoT can substantially alleviate the agriculture workforce's burden of data collection and sensor maintenance. This enables a greater concentration on critical agricultural tasks, ultimately leading to heightened efficiency and productivity.

Project Description: IoT will be with multiple types of sensors connected to monitor vital ambient environmental information of crops such as air temperature, humidity, soil moisture and nutrition level, and light intensity. The Collected data will be stored and pre-processed on IoT locally to get preliminary diagnosis of crops’ health. In particular, we will implement Tiny Machine Learning (TinyML) algorithms on IoT device to improve the performance and exert and benefits of the hardware accelerators on the IoT devices.

Student Research Activities: 1) Develop a sustainable IoT utilizing energy harvesting technology with commercially accessible electronic components. 2) Acquire proficiency in understanding the pinout of IoT devices and learn proper connections to the external sensors. 3) Design embedded algorithms for operating connected sensors. 4) Leverage onboard hardware accelerator to design and deploy data preprocessing algorithms using TinyML.

Project Background: Due to budget and power constraints, direct data transmission from IoT sensor modules on large farmland to a remote server is challenging. To address this, we propose using the emerging unmanned aerial vehicles (UAV) such as a quadcopter, quipped with storage, processing units, and communication modules to efficiently ferry data from ground IoT devices to edge servers. Additionally, UAVs with integrated cameras can capture morphological attributes of crops. Our purpose is to integrate multi-modality data from UAVs and ground-based IoT systems using machine learning to enhance our knowledge and understanding of environmental impact to the growth of crops and plants for better agricultural management.

Project Description: The lower-power IoT sensors on the ground will collect vital ambient environmental information of the crops such as air temperature and humidity, soil moisture and nutrition level, and light intensity. The UAV in the air will ferry ground IoT sensing information and capture morphological data such as color, flowering patterns, and canopy coverage. We will employ machine learning to integrate multi-modal data from both the IoT system and UAV to evaluate the healthiness of the crops and yield the best-effort crop management. Through this project, the student will learn to leverage emerging UAV technology and machine learning to understand crop growth and corresponding agriculture management.

Student Research Activities: 1) Enable UAV to receive data from IoT sensors and store the data into internal storage; 2) Design UAV’s flying path based on the distribution of ground IoT sensor; 3) Conduct feature extraction of crops’ morphological attributes from photos taken by UAV; 4) Learn to integrate IoT sensing data and extracted features from UAV photos using machine learning for spatiotemporal analysis of the health and growth of crops in the farmland.

Background: Understanding fruits during the growth stages is a critical step toward automated harvesting and yield prediction. Artificial Intelligence encompasses emerging state-of-the-art techniques that improves the ability to solve this problem and deal with the variations of the crop. We propose to research applying multispectral classification of tomatoes to evaluate the tomato growth stages and extracting information that include ripeness, size, and quantity that are important for automated harvesting systems.

Project Description: We will investigate the challenges and evaluate solutions to classify the characteristics of the tomato crop from its initial bud stage through to the full ripeness stage. This includes developing and utilizing multispectral sensors to collect data throughout the growth cycle (for example: spectral profile, size, and count). The gather data will be annotated and used to train a robust AI model. The data will also be provided to growth models to assess additional analysis that include shape characteristics for ripeness metrics. This will benefit students understanding the translation of crop analysis that will support harvesting and yield prediction. This will also demonstrate how AI tools are used in industry to better understand crop growth and manage them efficiently and effectively.

Student Research Activities: 1) Develop a data collection and analysis system to monitor a controlled set of tomato crops; 2) Train AI model tomato classification models and extract growth analytics; 3) Evaluate and benchmark the precision and recall for determining the tomato characteristics for harvesting.

Background: The distribution of fresh goods required human intervention for repetitive tasks such as quality inspection and placement which can be replaced by robots. A novel AI-powered automatic crop packaging, inspection, and assorting system can improve efficiency and reduce the need for human labor. Sponsored by a local grocery chain store, HEB, a senior design project successfully sorted meat packages into two bins using the prototype.

Project Description: The proposed system will improve the performance of the prior prototype with open-source AI for object detection and localization. It will inspect various types of fresh goods, directs a robotic arm to place them in appropriate bins, and establish a database to track the sorted packages considering the date, weight, and price.

Student Activities: 1) Expand the system to a larger working surface with multiple cameras for accurate package inspection and localization; 2) Improve the performance of the AI algorithm to identify more diverse fresh packages such as different vegetables and meats. 2) Refine the control algorithm for more accurate package placement with reduced power and improved efficiency.