Self-localization is an essential and widespread capability of modern smart phones. However, technologies like GPS usually only work in open spaces and are therefore not useful in indoor environments. In order to achieve fine-grained localization in indoor environments, additional sensors of smart phones (e.g., cameras, microphones, accelerometers, etc.) turn out to be more appropriate.
Since solving the localization problem using such sensors can be quite difficult due to different places sharing similar features, it is furthermore useful to actively gather information about the current location. While this strategy is commonly employed by mobile robots (e.g., driving around a corner in order to disambiguate two places), smart phones themselves do not possess any effectors to accomplish this. However, there might exit effectors in the environment with which the device can communicate (e.g., network-controllable lights like in the Cartesium). In this scenario, the device could form hypotheses about its current location by fusing information from its sensors, and, in case this does not lead to a unique state estimate, it could then try to influence the environment to improve the estimate.
The following problems will have to be addressed in this context:
- fusing noisy information from multiple sensors and estimating the current location
- interacting with environment effectors to improve the estimate
The resulting software should run under Android.
This topic is suited for a diploma or master thesis, which can be written in German as well as in English.
contact: Thomas Reineking
