UWB based high precision localisation technology for the low cost autonomous UAV inspection in GPS-denied and extremely confined environments

  • Beiya Yang

Student thesis: Doctoral Thesis

Abstract

Owing to the specific characteristics of unmanned aerial vehicles (UAVs), the demands and applications increase dramatically for them being deployed in extremely confined or closed space for surveying, inspection or detection to substitute human. However, global positioning system (GPS) may lose effectiveness or become unavailable due to the potential signal block or interference in such environments. Under such circumstances, an imperative requirement on new positioning technology for UAV has emerged. With the rapid development of radio frequency (RF) based localisation technologies, especially for the ultra-wideband (UWB) based localisation technology, leveraging small wireless sensor nodes for low cost, low latency, low energy consumption and accurate localisation on UAV has received significant attention. However, the research challenges and issues such as the unreasonable values within the UWB measurements, the geometry configurations of the anchor nodes in the extremely confined environments, the requirement of the prior information, the performance influence from the unpredictable propagation condition and the geomagnetic disturbances for inertial measurement unit (IMU) still exists which limit the applications on UAV. To avoid these aforementioned research challenges and issues, the researches in this thesis are carried out from different perspectives. Firstly, the maximum iii Chapter 0. Abstract likelihood estimation (MLE) based algorithm plus with the anchor distribution strategy is proposed focused on the development of the pure UWB based localisation system. With the proposed algorithm and strategy, the most suitable geometry configurations of the anchor node can be found to achieve the accurate and robust UAV positioning in focused environments. However, considering the unreasonable values within the UWB measurements still have the huge impact on the localisation performance for the pure UWB based localisation approach, according to the simulation and experiment, the investigation on the sensor fusion based approaches which integrated the IMU and UWB is carried out. To overcome the performance degradation and oscillation leads by the unreasonable values, the extended Kalman filter (EKF) based algorithm can be seen as the ideal candidate, owing to the implementation simplicity and acceptable accuracy. Yet, the unknown prior information about the noise covariance matrices still has the great impact on the localisation performance, especially for the applications in the extremely confined environment. To mitigate the effects, in this thesis, a high precision UAV positioning system which integrates the IMU and UWB with the adaptive extended Kalman filter (AEKF) is proposed. Compared with the traditional EKF based approach, the estimated and recorded information from previous processes is exploited to adaptively estimate and further control the estimation of the noise covariance matrices for performance improvement. Nevertheless, the proposed AEKF algorithm and system still suffer from the performance influence caused by the geomagnetic disturbances for the additional IMU. To remedy this for further performance improvement, in this thesis, the tightly coupled adaptive extended Kalman filter (TC-AEKF) based algorithm is proposed. With the additional angular rate in the state prediction process and iv Chapter 0. Abstract the adaptively estimated noise covariance matrices, the proposed algorithm can significantly improve the localisation performance in the focused environments, according to the simulations and experiments. Even the proposed AEKF and TC-AEKF based algorithm can attain the high accuracy and precision localisation performance of the UAV in focused environments, however, there is still limitation for the proposed algorithm. Due to the principal of the proposed algorithms, the additional linearisation process is required in the correction process. Currently, the first order Taylor expansion is utilised for the linearisation of the observation matrix. This may directly lead to the performance degradation and oscillation, owing to the neglected high order terms. To mitigate the effect, in this thesis, an adaptive square root cubature Kalman filter (ASRCKF) based sensor fusion algorithm is proposed. With the integration of the IMU and UWB, the utilisation of the cubature rule, the adaptively estimated noise covariance matrices and the added estimation weighting factors, the performance degradation and oscillation led by the unreasonable value within the ranging information, the linearisation of the observation matrix and the unknown and hard-to-adjust noise covariance matrices can all be resolved. Finally, from the numerical simulations, experiments and autonomous inspection flight test, it can be proved that the proposed algorithm and system can attain 0.081m median error, 0.172m 95th percentile error and 0.045m average standard deviation (STD), which is feasible for the autonomous inspection in the focused environments.
Date of Award5 Jun 2023
Original languageEnglish
Awarding Institution
  • University Of Strathclyde
SponsorsUniversity of Strathclyde
SupervisorErfu Yang (Supervisor) & Xiu-Tian Yan (Supervisor)

Cite this

'