Search for: All records

This paper presents a senior capstone design project to design a remote asset tracking and monitoring system platform by using an organization’s local network as a cost-effective alternative solution to a traditional global positioning system (GPS). The proposed system utilizes an existing local area network (LAN) infrastructure to train a machine learning (ML) model to predict and map the locations of an asset, such as a university shuttle. The proposed system was developed and implemented by using a Raspberry Pi single-board computer with an external Wi-Fi antenna to collect a pool of media access control (MAC) addresses and the relative signal strength indicator (RSSI) values as a fingerprint for each location. Furthermore, the latitude and longitude (GPS) data were captured at each of the collection points to train the machine learning models. Once the collected MAC, RSSI and logistical data, features were generated, processed, and exported to Elastic. To determine which model was suitable, we overlaid our chosen Decision Trees, Extra Trees, and Random Forest models on a map to visualize any deviations from our initial GPS data points. The results showed that the Decision Trees model performed the best, with most of the predicted points having an acceptable margin of error relative to our collected data. In the field-testing phase, the plan is to attach the prototype design onto a university shuttle to track its routes around the campus. The results will provide the feasibility of the proposed concept, and it will improve our community’s transportation needs by providing more efficient shuttle stops on campus. The long-term goal of the proposed collaborative research between Engineering, Computer Information Systems and Cybersecurity students is to provide safe and healthy spaces by integrating real-time indoor air quality (IAQ) data within the shuttles to support the community in making informed decisions on daily actions such as catching the shuttle timely on campus.