Robust Localization on Non-Planar Surfaces

Local positioning systems are valuable for a range of applications where the global positioning system (GPS) is unavailable or unreliable. In this paper work is presented on accurate tracking of athletes in indoor venues using time-of- arrival (TOA) measurements of radio signals exchanged between nodes and an iterative least-squares localization algorithm. The range error characteristics were measured and it is shown that contrary to the usual assumption the distribution is not a function of range, as sources of noise other than at the receiver input dominate except at very low signal levels. It is also shown that the range error does not follow a normal distribution, but has a higher probability of large errors which can result in large localization errors. An algorithm is presented that is robust to these errors, and results presented from measurements in a velodrome demonstrate the effectiveness of the algorithm. The cyclists are constrained to a track which is non-planar, thus two- dimensional localization is inaccurate, yet the number and distribution of nodes prevents accurate three-dimensional localization. An extension to the algorithm for accurate localization on a non-planar surface is also presented and shown to reduce the tracking error.