University students often face avoidable delays when navigating complex academic buildings and seeking reliable academic information, especially when classroom locations, accessibility routes, course prerequisites, and technical course content are distributed across multiple sources. This senior design project presents AUM Companion, a publicly deployed web application that integrates indoor wayfinding and a domain-restricted academic chatbot to support students at the American University of the Middle East. The navigation subsystem models the campus buildings as a multi-floor weighted graph derived from AutoCAD floor plans and laser-meter site surveys, with measurements recorded at 0.1 m precision. Dijkstra’s algorithm computes shortest paths across three floors, while elevator bridge nodes enable cross-floor routing. The chatbot subsystem applies retrieval-augmented generation using curated and administrator-approved Industrial Engineering course documents embedded in a Pinecone vector index. It answers course-specific and policy questions on prerequisites, technical concepts, worked examples, and curriculum relationships while refusing out-of-scope queries. Validation included algorithm benchmarking over 300 routing trials and chatbot testing on 70 complex student-realistic questions with 97.5% accuracy. Results showed optimal path generation with zero deviation from expected shortest-path cost and correct chatbot behavior for both in-corpus and out-of-corpus questions. The system reduces student wayfinding time, improves access to verified academic knowledge, and provides a low-cost, reproducible model for regional universities seeking practical digital campus-support solutions.

Keywords

Indoor Wayfinding, Accessibility, Academic Assistance, Retrieval-Augmented Generation, Dijkstra’s Algorithm