Compact, real-time localization without reliance on infrastructure

At Kaarta, we develop state-of-the-art technologies in real-time localization and mapping. While our key motivation is to enable robots to navigate reliably, and to separately build accurate three-dimensional representations of the environment. Our methods are suitable for localization applications since we use compact devices that can be easily hand-carried and do not rely on GPS. The Microsoft indoor localization competition provides us an opportunity to demonstrate two of our devices that use cameras, IMUs and laser scanners to produce highfrequency ego-motion estimation along with registered point clouds. We have successfully demonstrated our methods in both indoor and outdoor environments using data collected by us, as well as by others. In the latter case, our method is ranked# 1 on the KITTI odometry benchmark1. We see an average position error of approximately 0.2% of distance traveled, and we expect a better result in indoor settings.