Roomba Obstacle Mapping

Motivation

“The promise of IoT is smart everything” (ref). In fulfilling this promise for robotics, localization is the next frontier. And we wanted to achieve a solution without actual human presence. Our other motivation is to look for a different approach than current solutions available for indoor localization. WigWag is also working in this domain and was generous enough to sponsor our project.

Previous Work

When the team is researching on current presence detection technologies and mapping algorithms, many implementations involved in cameras to do the job. One typical implementation is using Xbox Kinect and SLAM algorithms to map the floor. It is very accurate and can show real life images. However, because of our hardware limitations, we do not have the option to use cameras in our project. Other implementations on localization can be achieved by using Radio Frequencies (RF). Techniques such as Fingerprinting, Maximum Likelihood localization, Sequence-based localization, or Ultra Wide Band-based localization. Also, during research, we found that people also trying to implement mapping using Roomba's internal odometry. However, they showed that Roomba's odometry was so inaccurate that it was almost impossible to do it.

    Roomba Odometry Roomba Odometry

Roomba Odometry Model Result from [1]

What We Contributed

We have seen research in RF-based localization and robot mapping, but we wanted to combine the two concepts into one simple, integrated solution that uses relatively cheap and readily available off-the-shelf consumer products. Using the iRobot Create 2, UC Berkeley's Tmote-Skys, and Estimotes Bluetooth beacons, we created a proof-of-concept robot to map the obstacles in a typical consumer's living room. By implementing every component as a Python module, other users can easily import our project, modify the algorithms, and test their code real-time. This provides people with a playground for scientific discovery and perhaps a starting point for WigWag to build a Home IoT solution for living room mapping.