Closing the Gap

Decentralized Structural-RNN for Robot Crowd Navigation with Deep Reinforcement Learning

“…For example, SeeWay uses skybox images to represent landmarks [20]. Similarly, Landmark AI offers semantic-related functionalities including describing the environment, reading road signs, and recognizing landmarks using a phone camera [19]. However, these phone applications are not robots and thus cannot physically guide users or provide a stable mounting point for cameras.…”

“…Thus, they are not built for PwVI that often need help in perceiving the environment and planning paths. Although attempts have been made to integrate vision or language models in wayfinding [11], [19], [20], how to build a guiding robot that can intuitively exchange semantic information with users remains an open challenge.…”

Wayfinding Tools for People With Visual Impairments in Real-World Settings: A Literature Review of Recent Studies

“…Travelers, at times, expressed an interest in holding the wayfinding device in their hands in order to receive real time haptic feedback during wayfinding tasks (Sato et al, 2019). Other participants desired a hands-free option during dynamic tasks (Abu Doush et al, 2016;van der Bie et al, 2019a;Saha et al, 2019). Still other research teams provided ways for participants to hold the phone in one's hands or release it on a lanyard or store it within a belt around their waists (Balata et al, 2018;Sato et al, 2019).…”

“…Across indoor and outdoor environments, participants remarked on the role of landmarks in supporting wayfinding tasks (Abu Doush, et al, 2016;Balata et al, 2016;Balata et al, 2018;Cohen and Dalyot, 2020;Guidice, et al, 2020;Saha et al, 2019;Sato et al, 2019). (Saha et al, 2019) described the saliency of the landmarks by having participants rank their preferences starting with most preferred: 1) Tactile landmarks, 2) Structural landmarks, 3) Sound, 4) Smell, 5) Air. (Cohen and Dalyot, 2020) applied these categorical descriptions of landmarks, when participants identified the useful landmarks along the routes through a park.…”

“…For example, some travelers wanted to find things that mobility tools don't normally find like railings, empty tables, and people blocking the pathway. (Saha et al, 2019). Although wayfinding technology is designed to look for the concise routes, visually impaired participants expressed that longer routes were perceived, at times, to be safer routes than the most efficient route (Zaib et al, 2019;Cohen and Dalyot, 2020).…”