Institutional Repository
Technical University of Crete
EN  |  EL



My Space

Unmanned vehicle navigation in GPS-denied environments

Apostolakis Georgios

Full record

Year 2020
Type of Item Diploma Work
Bibliographic Citation Georgios Apostolakis, "Unmanned vehicle navigation in GPS-denied environments", Diploma Work, School of Electrical and Computer Engineering, Technical University of Crete, Chania, Greece, 2020
Appears in Collections


This work summarizes the most important localization and simultaneous localization and mapping (SLAM) algorithms for indoor navigation and studies a RFID-based technique for improved localization accuracy. Stereo cameras and LiDARs are widely used to assist a robot’s movement inside a known or unknown environment, as they are not easily affected by the environmental noise. On the other hand, radio frequency identification (RFID) tags are very cheap and do not need any external power source, despite being affected a lot by the propagation environment. These characteristics have rendered them very popular and great research interest has risen in order to create accurate models for them. This thesis tests an approach which combines two measurements from each RFID tag, the received signal strength indication (RSSI) and the phase, in order to improve the localization/SLAM accuracy, compared to prior art. This integration is implemented by a particle filter, with an anchor tag being essential to compute the model parameters. For good estimation of the parameters, it is essential that the tags’ measurements are correlated, i.e. tags’ inter-distance should be less than half of the wavelength. Performance of the RSSI-phase integration was evaluated by comparing the estimation error to corresponding algorithms, which only use the RSSI or the phase. Localization error under light multipath (i.e. reflections of the radiation on surfaces of the area e.g. walls are combined and create a different than the expected signal to be received by the reader) was found in the order of less than 20 cm for all three algorithms, inside a 3D area of 12 m3 volume. However, the error of the algorithm which only used the phase, remained significantly higher (more than 40 cm) for some time after its beginning. Localization error under strong multipath was found in the order of 20 cm when only phase was used, 60 cm when only RSSI was used and 40 cm with their combination, inside a 3D area of the same volume with above. Again, the phase algorithm induced a great error (more than 1.5 m) before it converged. Therefore, the proposed approach offers a reduced localization error from the initial time steps and can be implemented in environments with both light and rich multipath.

Available Files