Real-Time Auditing of Domotic Robotic Cleaners

Domotic Robotic Cleaners are autonomous devices that are designed to operate almost entirely unattended. In this paper we propose a system that aims to evaluate the performance of such devices by analysis of their trails. This concept of trails is central to our approach, and it encompasses the traditional notion of a path followed by a robot between arbitrary numbers of points in a physical space. We enrich trails with context-specific metadata, such as proximity to landmarks, frequency of visitation, duration, etc. We then process the trail data collected by the robots, we store it an appropriate data structure and derive useful statistical information from the raw data. The usefulness of the derived information is twofold: it can primarily be used to audit the performance of the robotic cleaner –for example, to give an accurate indication of how well a space is covered (cleaned). And secondarily information can be analyzed in real-time to affect the behavior of specific robots – for example to notify a robot that specific areas have not been adequately covered. Towards our first goal, we have developed and evaluated a prototype of our system that uses a particular commercially available robotic cleaner. Our implementation deploys adhoc wireless local networking capability available through a surrogate device mounted onto this commodity robot; the device senses relative proximity to a grid of RFID tags attached to the floor. We report on the performance of this system in experiments conducted in a laboratory environment, which highlight the advantages and limitations of our approach.