State Estimation For Micro Aerial Vehicles Via Wireless-inertial Fusion

Over the last decade,the rapid proliferation of robotic technologies has made micro aerial vehicles(MAVs)fly into people's daily lives.Intelligent applications of MAVs are fundamentally supported by state estimation technologies.Currently,the dominated opticalbased state estimation systems are still lackluster due to the versatile nature of MAVs.They are limited in many environmental conditions,such as lighting,texture,and smoke.To address the above problems,this thesis proposes new state estimation technologies using radiofrequency(RF)signals as the environment perception medium.The diffraction,reflection,and penetration of RF signals make them immune to optical sensing limitations,thus improving the robustness of state estimation systems.Wireless-based state estimation is facing three challenges: 1)wireless localizability requires a costly pre-calibration of wireless anchors;2)wireless signals have a limited sensing range;3)wireless sensing suffers from an inferior accuracy.To surmount the above challenges,this thesis presents state estimation technologies that fuse wireless sensing to achieve initialization-free,remote sensing,and super accuracy.The main contributions are summarized as follows:1.To reduce the deployment cost for specialized hardware or wireless anchor calibrations,we present a multi-subcarrier-inertial fusion based MAV state estimation system.Our system is designed to be real-time and initialization-free,working upon existing wireless infrastructure(e.g.,Wi Fi)without any pre-calibration.It consists of two coupled modules.First,we propose an angle-of-arrival(Ao A)estimation algorithm to estimate MAV attitudes and disentangle the Ao A for positioning.Second,we formulate a Wi Fi-inertial sensor fusion model that fuses the Ao A and the odometry measured by inertial sensors to optimize MAV poses.The indoor experiments show that our system achieves the accuracy of MAV pose estimation with the position error of 61.7 cm and the attitude error of 0.92°,the accuracy of AP's position 53.4 cm.2.To address the limited sensing range of wireless signals in the use of state estimation,we present a chirp-inertial fusion based MAV state estimation system.It is nonintrusive to commercial MAVs by attaching backscatter tags to their landing gears without internal hardware modifications.It supports long-range/through-wall state estimation for MAVs by leveraging the high sensitivity of CSS signals.The enabling techniques are a new backscatterbased pose sensing module and a novel backscatter-inertial super-accuracy state estimation algorithm.We demonstrate our design by programming a commercial-off-the-shelf MAV to fly in different trajectories autonomously.The results show that Marvel supports navigation within a range of 50 m or through three brick walls,with an accuracy of 34 cm for localization and 4.99°for orientation estimation,outperforming commercial GPS-based approaches.3.To overcome the conundrum that optical-based solutions suffer from textureless scenes while RF-based ones have inferior accuracy due to their large wavelength,we present a UWBinertial fusion based MAV state estimation system.The system conquers the textureless challenge with RF-referenced monocular vision,achieving centimeter-level accuracy in textureless scenes.It consists of 1)an RF-sifting algorithm that maps 3D UWB measurements to 2D visual features for sifting the best features;2)an RF bias estimation algorithm that takes the sifted features to reversely estimate the RF bias for better accuracy;3)an RF-visual-inertial sensor fusion algorithm that leverages inertial sensors to combat the occasional loss of visual odometry.We implement the prototype with off-the-shelf products and conduct large-scale experiments.The results demonstrate that Sift is robust in textureless settings,10× more accurate than the state-of-the-art monocular vision system.In conclusion,this thesis addresses the challenges of MAV state estimation using wireless sensing and proposes an initialization-free approach,a remote sensing approach,and a super-accuracy approach.Combining all the approaches forms the new state estimation technology that uses wireless signals as a new sensing modality,being complementary to existing dominated vision-based approaches.All approaches have demonstrated the effectiveness by prototype implementations and real-world experiments.The proposed techniques can significantly improve the robustness of MAV state estimation.