BWIBots: A Platform for Robots Operating in Human-Inhabited Buildings
In February 2017, researchers at the University of Texas at Austin's Building-Wide Intelligence project published a description of the BWIBots platform in the International Journal of Robotics Research, detailing mobile robots designed to operate autonomously in office buildings, accept natural-language commands, and learn cooperative tasks through human interaction.
Background
Most robot deployments in the early 2010s followed a familiar pattern: a controlled environment, a limited task, and a short run. A robot might sort objects on a table, or navigate a cleared corridor, for the duration of an experiment. Then it was switched off. The gap between that kind of study and anything resembling daily use in a real building was wide.
The technical reasons were real. Office buildings are unpredictable. Lifts break, doors are left ajar, people walk across planned routes without warning. A robot that had mastered one floor plan in a lab would struggle the moment the furniture moved. Handling all of that required combining several difficult problems at once: navigation, perception, task planning, and communication with people who had no technical training and no patience for rigid command formats.
Natural-language instruction was its own challenge. Most systems of the time required users to choose from a fixed set of phrases, or to learn a syntax. Getting a robot to interpret a loosely worded request from an ordinary person, decide what the request actually meant, and then carry it out reliably, was not a solved problem. Neither was letting non-experts teach the robot new tasks without writing any code.
What happened
Piyush Khandelwal, Jivko Sinapov, Matteo Leonetti, Rolando Fernandez, Justin Hart, and Peter Stone, all working at the University of Texas at Austin’s Building-Wide Intelligence project, built the BWIBots platform to run inside the university’s computer science building continuously, not just during scheduled experiments. The paper describing the platform appeared in the International Journal of Robotics Research in February 2017.
The robots used ROS (the Robot Operating System, an open-source framework that handles the low-level communication between sensors, actuators, and planning software) as their base. On top of that, the team built interfaces that accepted natural-language commands, so a person could tell a robot to fetch something or escort a visitor without following a script. The platform also included a mechanism for learning cooperative tasks through interaction: a person could demonstrate a behaviour or correct the robot’s actions, and the system would update accordingly. This was meant to let the robots become more useful over time without requiring the user to be a programmer.
Running inside a real, occupied building for an extended period set this work apart from most contemporary robot research. The deployment produced data on how the system actually failed, how people actually spoke to it, and how the gap between laboratory conditions and everyday office life showed up in practice. The paper documented the architecture, the interaction methods, and the lessons drawn from that sustained operation, giving other researchers a detailed account of what it looked like to keep robots of this kind running outside a controlled setting.
Why it mattered
BWIBots demonstrated a sustained, multi-year deployment of autonomous service robots in a live office environment, combining natural-language understanding, task planning, and interactive learning in a single integrated platform. The project provided a concrete testbed for evaluating how robots can acquire new behaviours from non-expert users in unstructured human workplaces. Its long-running deployment distinguished it from short-duration laboratory studies, yielding real-world data on failure modes and human–robot interaction at scale.
People
Jivko Sinapov, Matteo Leonetti, Piyush Khandelwal, Rolando Fernandez, Justin Hart, Peter Stone
Organisations
University of Texas at Austin, Building Wide Intelligence Project Ut Austin
Sources
- BWIBots: A platform for robots operating in human-inhabited buildings.International Journal of Robotics Research (SAGE Publications).Primary source
- BWIBots: A Platform for Robots Operating in Human-Inhabited Buildings (author manuscript).University of Texas at Austin Department of Computer Science.institutional
- Building-Wide Intelligence Project – UT Austin.University of Texas at Austin Department of Computer Science.Official
Cite this page
AI Achievements. (2017). BWIBots: A Platform for Robots Operating in Human-Inhabited Buildings. Retrieved 2026-08-22, from https://achievements.ai/milestone/bwibots-the-visionary-robots
@misc{achievements_bwibots_the_visionary_robots,
title = {BWIBots: A Platform for Robots Operating in Human-Inhabited Buildings},
author = {{AI Achievements}},
year = {2017},
url = {https://achievements.ai/milestone/bwibots-the-visionary-robots}
}