Optimal landmark placement for indoor positioning using context information and multi-sensor data

The most effective solutions for indoor positioning of mobile agents typically rely on multi-sensor data fusion. In particular, good trade-offs in terms of accuracy, scalability and availability can be achieved by combining dead reckoning techniques (e.g. based on odometry) and measurements of distance and attitude with respect to suitable landmarks with a known position and/or orientation within a given reference frame. A crucial problem of this kind of techniques is landmark deployment, which should keep into account not only the limited detection range of the adopted sensors and the non-null probability of missing a landmark, even if it actually lies within the sensor detection area (SDA). This paper focuses on minimum landmark placement taking into account possible environment contextual information. This solution relies on a greedy placement algorithm that optimally solves the problem while keeping positioning uncertainty below a given limit. The correctness of the proposed approach is verified through multiple simulations in the context of the EU project ACANTO, which requires to localise one or more smart robotic walkers in large, public and potentially crowded environments such as shopping malls or airports.