Lessons Learned from the Development and Deployment of a Hotel Concierge Robot to be Operated in a Real World Environment

This paper documents the many valuable lessons learned during the development and deployment process of a hotel concierge robot to be operated in a real world environment, which are otherwise difficult to gain in a lab environment or from typical academic research. Real world requirements were explicitly stated by the client before the start of the project, and a significant time constraint of delivering the robot in 1 month was given. Despite the constrained timeline, a functional platform that met all the requirements was created and successfully deployed, where it is run for 24 hours a day. This was possible by relying on rapid prototyping, modifying off the shelf components, and creating a custom codebase based on a scripted language with open-source libraries. As the robot was to be operated in a real world environment, we have encountered numerous problems that are not often encountered in academic research. These include problems such as the robot being repeatedly "exposed" to human interaction, how clothes gradually prevent motions over time, keeping the robot up-to-date after the initial delivery, training of non-roboticists such that they can conduct basic maintenance, and issues that arise as runtime lengthens.