2026 ASEE Annual Conference & Exposition

Introducing STEM Students to Advanced IoT Design Concepts

The proliferation of Internet of Things (IoT) applications continues to increase the importance of effectively teaching engineering students about this technology. In support of this goal, an ongoing project at Texas A&M University-Kingsville and Texas A&M University-Corpus Christi has focused on enhancing the teaching of IoT concepts and technology. Throughout the project a particular emphasis has been placed on accommodating engaged remote student learners who might not have access to lab facilities, such as those with work or family obligations. A hands-on approach has been adopted in which students are tasked with completing lab exercises to learn about IoT concepts and technology. Students are provided with an IoT learning toolkit that contains a processor board along with sensors, actuators, resistors, LEDs, and the other elements needed to design and implement an IoT solution for each exercise. This enables students to work on exercises at a time and place that fits their schedule.

The exercises developed during the project have been organized into exercise sets, each focusing on a particular aspect of IoT technology. Students are instructed to complete the exercises within each exercise set sequentially since they progressively build upon each other in order to introduce and teach the associated IoT concepts. The first exercise set developed defined a series of tasks designed to introduce and familiarize students with basic IoT elements and how they can be combined to form an IoT solution. Subsequent exercise sets have expanded on this foundation to provide students with experience in designing IoT solutions that incorporate additional components and capabilities.

The latest exercise set developed during the project provides students with experience in integrating a camera module into IoT solutions. The exercises in the set demonstrate the use of the camera in the design of IoT applications for tasks in which image capture is required such as security surveillance or monitoring crops in an agricultural setting. The final exercise in the set also provides students with an initial experience in utilizing basic AI (artificial intelligence) capabilities in an IoT solution. Students are given the task of utilizing a cloud-based AI service (Google Gemini), accessed through an API (Application Programming Interface), to analyze an image captured by the camera.

The current paper reports on the new set of exercises developed to provide students with further experience utilizing image capture and AI technology in the design of IoT solutions. While the previous exercise set integrated cloud-based AI functionality, the new set focuses on deploying a machine learning (ML) model directly on the processor board itself and utilizing it for object recognition. Students are provided with support materials including an overview of the features and use of a specific pretrained machine learning (ML) model (YOLO12) that can run on the processor board provided in their IoT learning toolkit. The students are then asked to design IoT solutions for applications that require image capture and analysis, such as a simple security application that utilizes the camera to capture images at regular intervals, analyzes each image using the ML model, and raises an alarm if a person is detected in an image. The use of object recognition in this application ensures the alarm will only be triggered by a person and not an animal, thus being more accurate than a simple motion sensor.

The tasks that make up the new exercise set are designed to expand student knowledge in important ways. Utilizing a ML model running on the local processor board rather than accessing a cloud-based service through an API will introduce students to the edge-based approach of designing IoT solutions in which as much processing as possible is to be done as close to a data source as possible. Additionally, students will also learn about utilizing tools such as OpenCV (Open Source Computer Vision) and Numpy (Numerical Python) to facilitate interacting with a locally running ML model directly rather than accessing a cloud-based AI service.