On aspects of indoor localization

Abstract

One of the key issues of emerging mobile computing and robotics is to obtain knowledge of the position of persons, vehicles and objects in an indoor environment. In order to enable a pervasive coverage, this must be achieved at low cost. Due to the importance of this problem, numerous solutions have been proposed. In this thesis two important localization techniques are discussed and major improvements are developed to address their weaknesses.

A common approach for the position estimation problem is hyperbolic localization. This method is based on time difference of arrival (TDOA) measurements of signals transmitted from the mobile unit, i.e. the unit that is to be located, to a set of fixed reference units. A key requirement of this technique is that the clocks of the reference devices are synchronized. Unfortunately, if electromagnetic signals are used that travel at the speed of light even a very small clock deviation results in a tremendous inaccuracy of the computed location. Therefore, a highly precise time measurement and synchronization is mandatory for these systems. However, clocks operating at the required resolution and precision are complex and thus expensive. In this thesis, a localization approach based on TDOA measurements that implies clock synchronization is proposed and discussed. It allows for the usage of low cost oscillators that operate at a moderate frequency in the order of 100 MHz. The impact of short and long term instabilities like jitter or clock drift are inherently delimited. The approach can be applied to radio or infrared as well as ultrasound based systems. The main advantage of this approach is that common synchronization mechanisms that require a significant amount of processing and/or hardware resources can be neglected. Hence, this method is well suited for applications where the mobile unit must be very low cost and thus, of low complexity.

An alternative to hyperbolic localization is triangulation, which requires the measurement of angles among fixed reference units and the mobile target. In this work, an infrared detector array for angle of arrival measurement is presented. The array consists of multiple sensing elements that are orientated in different directions. First, the arrangement is described by a sampling system. However, it is shown that low cost integrated receivers yield various aberrations, and thus the sampling approach fails. Subsequently, a new approach named virtual filter interpolation is proposed and discussed. This approach can handle individual sensitivity characteristics of each sensor element and therefore outperforms the sampling approach. The proposed technique is qualified for extremely low cost localization, e.g. for robot navigation.