Navigation in GNSS denied environments using software defined radios and LTE signals of opportunities.
Thesis DisciplineElectrical Engineering
Degree GrantorUniversity of Canterbury
Degree NameMaster of Engineering
The aim of this project is to implement a positioning system based on existing technolo- gies as a backup for GNSS to be used in an Unpiloted Aerial Vehicle (UAV) designed for carrying passengers where a backup is a necessity. The project sponsors have set a target accuracy of <100 m.
A comprehensive list of existing and potential positioning technologies were assessed and LTE was selected as the most suitable, mainly based on accuracy, coverage, and availability. Other factors such as size, weight, power, and cost were also considered. LTE is a 4th generation cellular technology, used worldwide, and likely to be around for another decade. It presents the best accuracy based on current literature with  claiming GPS level accuracy.
LTE has two methods of positioning that do not require any changes to the network, 1) using Timing Advance (TA) and, 2) using LTE Signals Of Opportunities (SOP). Both were assessed; first the TA method, then the SOP method.
The TA value is a measure of the round trip time of the LTE signal from a Mobile Station (MS) to the Base Station (BS) and back. It has a granularity of 78 m, and hence, a theoretical positioning accuracy of 78 m. A Ublox cellular module was used to assess the suitability of TA for positioning. A practical accuracy of 300 m (95 % probability) with an update rate of 20 s per update was achieved. This result does not meet the accuracy criteria and its update rate of 18 s/update is lower than desired.
The SOP method tracks the Time-Of-Arrival (TOA) of LTE pilot sequences (i.e., signals of opportunities) and uses them to get the distance between the MS and BS. By getting distances to four BSs, the MS can uniquely identify its position. The pilot sequences were tracked using a LTE software defined radio made up of a USRP B200mini radio front-end and srsUE (an open source software implementation of an LTE MS written in C and C++). The resolution of the TOAs is one sampling period (Ts); which at 23.04 MHz equates to a positioning granularity and theoretical accuracy of 13 m.
In practice, the TOAs were detected to a resolution of 130–254 ns (95 % probability) depending on the environment. Both results were obtained from tests carried out at ground level where multipath was present. This result leads to a practical accuracy of 39–76 m (95 % probability) respectively, although this does not take into account time drift. To detect TOAs with sub Ts resolution, the Blais-Rioux peak detector algorithm  can be used. A possible method of removing multipath is provided in .
The biggest source of error in the system was the MS clock’s time drift (i.e., its time drifting away from real time). Hence, a practical and reliable positioning system will have to use a higher quality clock and/or a robust method of modelling and compensating its drift. This is difficult with low quality clocks such as the B200mini’s Temperature Compensated Crystal Oscillator (TCXO), whose drift was found to be non-linear and volatile; changing from trial to trial. Hence a higher quality clock is recommended, since its drift rate is smaller and also linear over a much longer time, making it easier to model and compensate.
A list of even higher quality clocks (for the MS and BS) are also provided, which would enable the time drift of the overall system to be negligible for the entire flight of the UAV, i.e., for ∼60 minutes after GNSS signal loss. Upgrading BS clocks is likely infeasible. The Spark networks current BSs are equipped with high precision Oven Controlled Crystal Oscillator (OCXO)  and disciplined by GPS; which means that their drift is constantly corrected, and cannot be observed and modelled for drift compensation. Without drift compensation a BS clocks time drift will likely exceed an error of 100 m in 5–10 minutes, which may only be suitable for emergency landings.
Overall, the LTE SOP method provides satisfactory positioning results, but a fully implemented practical and reliable system will need more research and development and system components are likely to be relatively expensive.