Unmanned aerial vehicle applications require trajectory information for planning and tactical operations. Initialised by positioning, navigation, timing technologies such as global navigation satellite systems, trajectory prediction models are used to determine future states or trajectories. Currently, research on unmanned aerial vehicle trajectory prediction is limited focusing on a subset of applications, without a universally agreed set of key performance indicators against which all the elements of unmanned aerial vehicle operations are evaluated. Another limitation is that current trajectory prediction models do not support all known location-based unmanned aerial vehicle applications and all aspects of their operations. To address these limitations, this thesis develops a new generic trajectory prediction model applicable to all known applications and operations.
Firstly, all location-based unmanned aerial vehicle applications and current regulations/specifications are reviewed and consolidated in exhaustive lists to reflect their relationships. This is followed by definition and specification of the key performance indicators, on the basis of which the architecture required for unmanned aerial vehicle operational functionality is specified. To underpin this functional architecture, state-of-the-art systems are reviewed and used to specify an end-to-end physical architecture to support operations. The thesis then focuses on the aspects of this architecture that require trajectory prediction, firstly by using the performance metrics derived from operational- and system-level requirements for each operational element, to define and specify the performance indicators for trajectory prediction. The indicators are then used to develop the functional architecture for trajectory prediction and its underpinning models, suitable for all applications.
Currently, there is a wide variety of trajectory prediction models that support a subset of applications. Regardless of the differences between these models, trajectory prediction accuracy is the fundamental performance measure and is the focus of this thesis. Following a review of the literature, this thesis specifies mathematical methods for trajectory prediction error budgeting and develops a more accurate trajectory prediction model than the state-of-the-art, informed by new model-driven methods and physics-based models for error mitigation. Compared with the state-of-the-art in trajectory-location-routing models which suffer from the limitation of weak performance modelling, the developed trajectory prediction model incorporates novel generic unmanned aerial vehicle performance and wind models to increase trajectory prediction accuracy in real-world environments.
The new trajectory prediction model proposed in this thesis is validated by three case studies in relation to the most stringent location-based applications. The validation process is implemented and performance in terms of accuracy is quantified. The results show that the new trajectory prediction model meets the accuracy requirements (i.e., 5m) of the most stringent applications (i.e., humanitarian delivery logistics, urban passenger transport), and achieves a higher accuracy than the current models. In the new model proposed in this thesis, trajectory prediction accuracy is improved by at least 1.55m compared with other state-of-the-art models. This improvement benefits all stakeholders. Firstly, it advances knowledge in consolidating and enhancing unmanned aerial vehicle trajectory management literature facilitating further research in relevant aspects of unmanned aerial vehicle development and applications, and trajectory-based operations. Secondly, it enables regulators to specify the necessary operational framework underpinned by credible safety assessment. Finally, it enables operators and service providers in strategic, pre-tactical and tactical operations of all trajectory-based operations to enhance operational capacity due to improvement in trajectory prediction accuracy.