The need for robust and accurate indoor localization solutions on smart phones and other connected devices has grown substantially in recent years. Consumer tracking, localized augmented reality, and navigation assistance are examples of the numerous emerging applications. Solutions proposed up to this date require either time-intensive calibration of the environment, dedicated hardware, or the active visual scanning of known landmarks or pre-registered maps. We are working on “blind” localization solutions which require no or little prior knowledge of the environment and yield sufficient localization accuracy. We believe that the future of localization is going to be “multi-modal”, combining:
- distance, angle and distance-difference measurements,
- more than one sensor types,
- fixed and variable anchor positions.
Motivated by the fact that angular measurements can be an excellent complement to range, for example when depth sensors are unavailable or fail, we am working on more optimal ways to include angular information in localization. As a first step, we are working on establishing a consistent theory for angle-based localization methods. Most state-of-the-art methods using angular information are suboptimal: some exhibit unwanted behaviors such as noise propagation, others do not allow the inclusion of other measurement types such as distances. The noise models are barely adapted for the periodic nature of angle noise, and directed vs. non-directed angles are not consistently differentiated.