Synchronisation et positionnement simultanés d'un réseau ultra-large bande et applications en robotique mobile

Mobile robots require accurate real-time location estimates to operate. These estimates use external measurements, obtained for example from Global Navigation Satellite Systems (GNSS), to correct integration errors from proprioceptive sensors. This master's thesis fo-cuses on short-range Ultra-Wide Band (UWB) radios as a source of external range mea-surements, which can be used to provide centimeter-level positioning accuracy in indoor environments, where GNSS signals are generally unavailable. UWB-aided positioning systems most commonly rely on pseudorange measurements obtained by multiplying the speed of light by the Time-of-Flight (ToF) of messages transmitted be-tween UWB nodes with known positions, called anchors, and UWB nodes to localize, called tags. To achieve the localization accuracy desired for indoor applications, errors in ToF mea-surements need to remain below the nanosecond. The main challenge in achieving this level of accuracy comes from the fact that internal clocks at di˙erent nodes are not synchronized. Simple two-way ranging protocols can provide ToF measurements without synchronizing the nodes, but require the tags to transmit messages to the limited number of anchors and as a result do not support more than a few tags. Hence, motivated by applications requiring the deployment of multi-robot systems, we focus on one-way ranging protocols with the tags operating as passive receivers, which however requires synchronized anchors. The main con-tribution of this thesis is to design a protocol based on Kalman filtering to simultaneously synchronize UWB nodes and obtain ToF measurements between active anchors and passive tags. The modeling and rejection of outliers in these measurements is discussed in details in order to improve the robustness and performance of the proposed algorithm. The pseu-dorange measurements are also fused with other sensor measurements to design UWB-aided integrated navigation systems. Our algorithms are implemented on a custom embedded platform combining a commercial o˙-the-shelf micro-controller and UWB radio with inertial sensors, and interfaced with a standard software framework for robotics. We characterize the localization performance achievable in practice through several indoor experiments with ground robots.