Self-Localization and Interpersonal Proximity Detection for Public Transit Accessibility and Safety
- Author(s): Mirzaei, Fatemeh
- Advisor(s): Manduchi, Roberto
- et al.
Public transit stations and hubs are difficult to navigate for people with visual impairments. Moreover, public transit has been affected disproportionately by the social distancing requirements consequent to the COVID-19 pandemic. It is the objective of this dissertation to provide a technology for addressing these concerns in the frame of a mobile app named RouteMe2. The technology provides micro- routing and guidance to visually impaired travelers through complex routes in transit hubs. This work also includes the study to monitor the distance between the travelers inside the bus for social distancing application. Reducing the risk of air-born viral infections by social distancing can contribute to improving the overall safety of the public transit. The key enablers of this technology are sufficiently accurate self-localization and micro-routing as well as effective communication of the contextual spatiotemporal information with the visually impaired users. The accuracy of the self- localization in the outdoor environments is challenged by poor Global Positioning System (GPS) reception due to tall nearby buildings that may obscure view of one or more satellites — a.k.a shading. Shading is very common in urban environments, and is a major cause of GPS failure. In order to mitigate the effect of shading, I statistically fuse the signals received from GPS as well as a small number of Bluetooth Low Energy (BLE) beacons. I further pair the statistical fusion with a Bayes discrete filter tracker to increase the self-localization accuracy. Experiments were conducted at San Jose Diridon light rail station to quantitatively assess the performance of the resulting system. I have designed and implemented certain features and functionalities of RouteMe2 to provide effective communication of the in-context spatio-temporal information with visually impaired users while they use the app. I leveraged our previously published focus group study conducted with visually impaired people as well as reviewing the user interface of the existing related apps to improve the user experience of RouteMe2 the detail of which is presented. I further assess the ability of two RSSI-based methods at detecting interpersonal distances shorter than 1 or 2 meters. One method uses the power received from the smartphone carried by another person. The other method measures the disparity in the power received by the two smartphones from one or more fixed BLE beacons. The results show that use of the RSSI disparity enables discrimination measures that are as good or better than using the RSSI received from another smartphone. I demonstrate the potential of a system that uses BLE beacons, placed inside a vehicle, to localize a passenger within the length of the vehicle with an accuracy better than 1 meter.