Pedestrian localisation for indoor environments

Ubiquitous computing systems aim to assist us as we go about our daily lives, whilst at the same time fading into the background so that we do not notice their presence. To do this they need to be able to sense their surroundings and infer context about the state of the world. Location has proven to be an important source of contextual information for such systems. If a device can determine its own location then it can infer its surroundings and adapt accordingly. Of particular interest for many ubiquitous computing systems is the ability to track people in indoor environments. This interest has led to the development of many indoor location systems based on a range of technologies including infra-red light, ultrasound and radio. Unfortunately existing systems that achieve the kind of sub-metre accuracies desired by many location-aware applications require large amounts of infrastructure to be installed into the environment. This thesis investigates an alternative approach to indoor pedestrian tracking that uses on-body inertial sensors rather than relying on fixed infrastructure. It is demonstrated that general purpose inertial navigation algorithms are unsuitable for pedestrian tracking due to the rapid accumulation of errors in the tracked position. In practice it is necessary to frequently correct such algorithms using additional measurements or constraints. An extended Kalman filter is developed for this purpose and is applied to track pedestrians using foot-mounted inertial sensors. By detecting when the foot is stationary and applying zero velocity corrections a pedestrian’s relative movements can be tracked far more accurately than is possible using uncorrected inertial navigation. Having developed an effective means of calculating a pedestrian’s relative movements, a localisation filter is developed that combines relative movement measurements with environmental constraints derived from a map of the environment. By enforcing constraints such as impassable walls and floors the filter is able to narrow down the absolute position of a pedestrian as they move through an indoor environment. Once the user’s position has been uniquely determined the same filter is demonstrated to track the user’s absolute position to sub-metre accuracy. The localisation filter in its simplest form is computationally expensive. Furthermore symmetry exhibited by the environment may delay or prevent the filter from determining the user’s position. The final part of this thesis describes the concept of assisted localisation, in which additional measurements are used to solve both of these problems. The use of sparsely deployed WiFi access points is discussed in detail. The thesis concludes that inertial sensors can be used to track pedestrians in indoor environments. Such an approach is suited to cases in which it is impossible or impractical to install large amounts of fixed infrastructure into the environment in advance.