High-precision position estimation in indoor environments

Abstract

I survey the state of the art on high-precision position estimation in indoor environments with a focus on our own work at the Freie Universitat Berlin. Indoor position estimation methods are introduced and shown to be sensitive to underlying assumptions on errors distributions. Performance of algorithms is shown to be location dependent. We compare results obtained by simulation to results obtained by experiments. Differences are caused by wrong assumptions on errors in simulation. Errors in experiments are less random than assumed, but proper models are too complex to be practical. To evaluate our experiments, we built a testbed and explain our robot based solution using off-the-self components. Ground truth is obtained using map matching, odometry, and scanning with a Kinect. The precision of 6.7cm is sufficient for current approaches with show errors one magnitude larger.