Student Builds AI Guide for Indoor Navigation
A Kennesaw State student is developing a voice-driven system that helps people find their way inside buildings without needing specific room numbers.
For Robert Hoeh, a software engineering student at Kennesaw State University, the struggle of getting lost in a large, unfamiliar building could soon become a thing of the past. He is developing an augmented reality navigation system that allows users to simply ask for directions in plain language. Instead of memorizing room numbers or reading signs, a user can speak a request like 'take me to the nearest bathroom,' and the system will guide them there. This approach aims to make indoor wayfinding more intuitive and accessible.
The project, reported by GN technics/ai (en-US), leverages consumer hardware such as the Meta Quest headset to create a hands-free experience. While traditional maps often rely on static labels, this system operates more like a conversation. It processes natural language requests, searches for the best match in available location data, and provides step-by-step guidance. The goal is to reduce the cognitive load associated with navigating complex indoor environments, particularly for those who may find such tasks overwhelming.
Designing for accessibility and safety
The primary motivation behind the project is to support individuals with cognitive disabilities and first responders. Navigating an unfamiliar structure during an emergency or for daily routine tasks can be stressful and confusing. By removing the need to identify specific locations beforehand, the system lowers the barrier to entry for wayfinding. Hoeh believes that a conversational interface is more forgiving than traditional navigation apps, which often require precise inputs and offer limited flexibility when a user is unsure of their destination.
Technical challenges in indoor mapping
Developing this technology is not without significant hurdles. Unlike outdoor navigation, which relies on satellite signals, indoor positioning requires different methods. Hoeh’s team has had to debug complex code and refine the accuracy of the system’s responses. Early prototypes were unreliable, sometimes failing to interpret basic requests correctly. This required extensive testing and troubleshooting to ensure that the AI agents could accurately interpret user intent and select the appropriate destination from the database.
The system was tested in the Atrium Building on the university’s Marietta Campus. During these trials, the team focused on how the AI handles ambiguous requests. If a user’s query is unclear, the system is designed to ask for clarification rather than guessing incorrectly. This interactive element is crucial for building trust and ensuring that users receive reliable directions. The development process has been a learning experience for Hoeh, who joined the research effort during his first year of university through a summer program.
Future steps for the project
The next phase of the project involves completing the system and conducting formal user studies. These studies will measure how effectively the technology serves its intended audiences, including students, staff, and potential first responders. Hoeh, who is entering his second year of study, has already prepared a paper on the work, which is nearing publication. His advisor, Brooke Zhao, notes that his persistence in overcoming technical bugs has been impressive. The project stands as an example of how undergraduate research can yield practical tools that address real-world accessibility challenges.
