VisionWay Accessibility

Existing solutions include cognitive maps which can help PVI confidently walk in known locations but need to be built on experience, and thus cannot be used to determine the best path towards an unfamiliar location. In fact, previous research has found that PVI, to reduce stress and increase safety, before embarking on a new journey need to be informed about whether a specific path has a paved sidewalk, street crossing point, controlled or uncontrolled crosswalk, stairs, obstacles, and potholes. Knowing these characteristics ahead of time, i.e., before starting to walk, can help PVI make choices on which path to take, which is particularly beneficial in unfamiliar environments. Trained Orientation and Mobility (O\&M) specialists often provide high-quality in-person help for PVI to walk in unfamiliar areas, notifying them of upcoming obstacles and turns on the path. However, a shortage of qualified specialists has increased the demand for remote instructors by leveraging smartphones as pointing devices. For example, services such as TappyGuide and WeAssist work by sharing the PVI's destination as well as smartphone camera and GPS data with a remote human operator, who can then consult a map to voice-guide the PVI to their destination. On the other hand, the remote directions depend on the availability of the remote operators. To alleviate this dependency, new technologies leverage mobile computing and Artificial Intelligence (AI) to sense the environment for obstacles and provide real-time navigation directions through audio or haptic feedback. For example, WeWalk's smart cane provides basic sensing and smart features to support navigation and obstacle detection alert; as part of USDOT's Inclusive Design Challenge, MPI Giudice, MPI Doore, and Co-I Fink developed the Autonomous Vehicle Assistant (AVA) app, which helps PVI safely walk towards an autonomous vehicle and accurately reach its door handle. Some recent projects target indoor localization and navigation or leverage augmented reality and haptic feedback to aid obstacle avoidance during navigation and reduce veering. While most funded projects focus on providing real-time navigation guidance given a certain path to follow, the reality is that more than one path may exist to reach a certain destination, especially in an outdoor context. Thus, they cannot help PVI and remote operators choose the best path to follow, which requires creating accessibility maps. The problem of creating accessibility maps has been widely studied for people in wheelchairs, whose main challenge is to ensure availability of accessible entrances or curbs on the path. WheelMap is an example of such a tool, but it does not help PVI, whose definition of accessible path can be quite different. For example, a wheelchair user may find it less problematic to handle an uncontrolled zebra crossing than someone with profound or total vision impairment. The most related solution focusing on finding accessible paths for PVI is that of Cohen and Dalyot, which leverages accessibility information already available in OpenStreetMap to evaluate the accessibility of different paths for totally blind users. However, it has three limitations. First, OpenStreetMap, similar to Google Maps and Apple's Maps, does not provide important information about many harmful and beneficial sidewalk features affecting accessibility such as tactile tech, potholes, tree branches, or sidewalk restrictions. Second, it remains unclear how to describe each path's features to PVI for selection. Third, it only focuses on blind pedestrians while the definition of accessibility and the features affect it can vary widely across severe, profound, and totally blind PVI.The goal of this project is to involve PVI with different impairment severity and mobility aids in the development of VisionWay, a mobile/cloud framework that helps PVI select a pedestrian path towards their destination based on three main innovations. This study focuses on the first of these by utilizing a user study designed to reveal how various sidewalk features affect accessibility and how to design an accessibility model able to estimate it. To address this challenge, barrier, and achieve our goal, the investigators assembled a team of men, women, trainees, and members who have lived experience with vision impairment (VI). Our strengths include engineering, computer science, human factor engineering, mixed methods research, human-computer interaction for PVI, clinicians including low vision and orientation and mobility specialists (OMS), as well as team members and participants with VI. The investigators' goal is to develop a mobile live, local, outside map framework called VisionWay, customized for PVI-user as a wayfinding tool to analyze, before starting to walk, the accessibility of candidate paths and choose one to the destination. The investigators designed aims for R61 technology development and feasibility. (Aim 1) Collect survey data and observations in pilot pathways to define and validate a PVI-facing metric to measure and detect the accessibility of a walk path from static and dynamic imagery.