Crowdsourcing the Installation and Maintenance of Indoor Localization Infrastructure to Support Blind Navigation
Cole Gleason, Dragan Ahmetovic, Saiph Savage, Carlos Toxtli, Carl Posthuma, Chieko Asakawa, Kris M. Kitani, Jeffrey P. Bigham · 2018 · Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies (IMWUT) · doi:10.1145/3191741
Summary
This paper introduces LuzDeploy, a physical crowdsourcing system that coordinates non-expert volunteers to install and maintain Bluetooth Low Energy (BLE) beacon infrastructure for indoor navigation systems used by blind people. The central problem is that indoor navigation systems like NavCog require dense networks of BLE beacons placed throughout buildings, and the installation and ongoing maintenance of this infrastructure is expensive, requires expertise, and degrades over time as batteries die or beacons go missing. LuzDeploy addresses this by breaking the complex installation and maintenance process into simple micro-tasks that untrained volunteers can complete with step-by-step guidance from a Facebook Messenger chatbot. The system has three components: the LuzDeploy Bot (a chatbot that assigns and guides tasks), the LuzDeploy Map (a web interface for administrators to plan beacon placements on building floorplans), and the LuzDeploy Data Sampler (an iOS app for collecting Bluetooth signal fingerprints). Tasks include placing beacons at designated locations, collecting Bluetooth signal samples for building localization models, sweeping hallways to check beacon health, and replacing broken beacons. The system was evaluated through a real-world deployment across four sessions over several months in a 7-story building at Carnegie Mellon University, with 127 total participants completing 455 tasks.
Key findings
Across four deployment sessions, 127 non-expert participants successfully installed 196 beacons and collected 1,823 Bluetooth signal samples in a 7-story university building. In Session 1, 89 volunteers placed 99 beacons in a single afternoon event, with 73 (74%) placed correctly within 1 meter of the intended location. Session 2 improved accuracy to 83 of 97 (86%) placed correctly over four months with just 25 participants, suggesting that fewer but more engaged volunteers produce better results. The crowd-built localization model achieved a median error of 1.6 meters compared to 1.2 meters for the expert-built model using LiDAR — only a 0.4 meter increase in median error, which still provides usable navigation accuracy. The total cost of compensating crowd workers for Sessions 3 and 4 was under $300, significantly less than hiring expert installers. The study identified key challenges: volunteers struggled with ambiguous map locations, placed beacons on incorrect surfaces (e.g., glass instead of walls), and inconsistently followed height guidelines. Altruism was the strongest initial motivator (making the building accessible for blind people), while competitiveness via leaderboards and monetary incentives helped sustain engagement over time.
Relevance
LuzDeploy tackles one of the most significant barriers to widespread adoption of indoor navigation for blind people: the cost and complexity of installing and maintaining the physical infrastructure. While the technology for indoor localization exists and works, most deployments remain confined to research settings because no one maintains the beacons after the initial installation. The crowdsourcing approach demonstrated here offers a scalable alternative — the 0.4 meter accuracy trade-off compared to expert installation is modest and still produces usable navigation. For accessibility practitioners and building managers considering indoor navigation systems, this research provides practical evidence that volunteer-maintained infrastructure is viable and significantly cheaper than expert installation. The findings about volunteer motivation are particularly useful: altruism works for recruitment but not retention, and task design matters enormously — batching tasks, providing clear landmarks rather than abstract map coordinates, and allowing flexible scheduling all improved outcomes. The work also raises important questions about long-term sustainability and whether crowdsourced maintenance can keep pace with infrastructure degradation in real-world buildings.
Tags: indoor navigation · crowdsourcing · Bluetooth beacons · blind users · wayfinding · infrastructure maintenance · physical crowdsourcing