Research Questions

How do you address the current capacity shortages of air travel in European airspace by increasing the overall level of automation of ATM, whilst maintaining or enhancing safety and minimising the impacts of aviation on the environment?

Can a trajectory prediction tool be developed with the necessary integrity to provide sufficient confidence to controllers and pilots for their use in a real time environment?

Research Scope

The aim of TESA (Trajectory prediction and conflict resolution for En-route-to-en-route Seamless Air Traffic management) was to develop reliable Trajectory Prediction (TP) and Conflict Detection and Resolution (CDR) capabilities with the specific objectives to address the sources of error (uncertainty) in TP and to use the improved TP to optimise CDR (thereby enhancing safety). The TP and CDR models were to be validated with real operational aircraft data and sensitivity analyses undertaken to characterise performance.

During the course of the project, emphasis was placed on the development of the TP tool. As a result, the scope of the development of the CD tool was limited, using simulated data only. CR research comprised a high-level study of conflict resolution methodologies based on the new TP and CD models.

Research Results

TESA developed a TP tool capable of predicting trajectories gate-to-gate in advance of the operation. This was achieved by extending Imperial College London’s en-route TP capabilities to the TMA and taxiing over time horizons of any duration under ideal conditions. The main innovations here were new models to reduce the ambiguity and complexity of aircraft intent representation for complex manoeuvres, accounting for aircraft dynamics and operational limitations (e.g. performance limits) as well as the wind impact. For taxiing, a simple model of surface friction was developed. All models were developed on the basis of real data from flight-data records. Under ideal conditions, the enhanced tool achieved maximum absolute errors with respect to a real trajectory of 2nm (along-track), 36 seconds (along-track), 0.32 nm (cross-track) and 575ft (altitude) over a look-ahead time of approximately two hours.

In order to account for the various sources of uncertainty in trajectory prediction, TESA developed the capability to simulate TP uncertainties and carried out a sensitivity analysis on the principal sources to determine those that require further modelling. The results from the sensitivity analysis showed that the largest sources are the uncertainties in the initial mass, as well as the along-track and cross-track components of wind. The impact of the control uncertainties was found to be most significant during the TMA phase of flight due to the significantly more complex manoeuvres.

The simulated uncertainties were used to develop a predictive TP uncertainty model distinguishing between two primary error sources: model and input data. TP model errors were developed on the basis of real Flight-Data-Record (FDR) data as a function of aircraft operational state. Input data errors were taken from relevant data sources where available and, in the absence of relevant documentation, assumptions were made on the remaining input data errors. TESA developed a strategy for combining these two error sources to obtain a conservative estimate of predictive TP uncertainties.

TESA developed a performance evaluation strategy, including its uncertainty modules. The evaluation strategy focussed on both short- and long-term predictions, with look-ahead times covering the entire duration of a flight. An important finding was that the accuracy of the TP 4D state-parameters is not sufficient to guarantee a representative aircraft configuration: the same trajectory can be achieved with different aircraft configurations. TESA thus iteratively optimised the TP model by minimising 4D position errors as well as aircraft control parameter errors (e.g. thrust and drag), speed and mass errors. While the 4D state parameters (longitude, latitude, height and time) are the primary parameters of TP accuracy performance, the prediction performance of the remaining parameters (referred to as secondary parameters) is also important. These secondary parameters are a direct reflection of the level of realism with which the trajectory has been simulated. Another key performance metric assessed by TESA is TP integrity, a measure of the level of confidence that can be placed in the tool. The performance was evaluated using real FDR data, comparing the predicted parameters under ideal and error-conditions from the TP engine with the observed parameters from the FDR data. The results show that the currently proposed accuracy performance targets for Time-Of-Overfly over a given point (set at 30 seconds – 95% – for the en-route phase of flight) can be met. However, the target of 10 seconds (95%) for the TMA phase of flight is unlikely to be met.

TESA developed a sophisticated Conflict Detection (CD) model that takes into consideration the uncertainties associated with TP, which are predicted ahead of time. This enables not only short-term conflict detection, but also the detection of conflicts over much larger volumes and time-horizons, thereby adding a holistic dimension to conflict detection. TESA’s approach predicts the instantaneous uncertainty volume at each point in space and time, ahead of time, accounting for the predicted TP uncertainties separately in each dimension. This volume is dynamic in all dimensions and representative of the predicted contextual conditions prevailing at the time at which the aircraft is predicted to pass the given point. It thereby enables a realistic and optimal detection of conflict risk, improving not only accuracy but also reliability and robustness.

TESA developed a high-level strategy for conflict resolution with the goal of taking stakeholder preferences into account as well as maximising safety and minimising the environmental impact of trajectory based operations. Key novelties enabled by the TESA developments are: improved prioritisation logic on the basis of actual collision risk during conflict resolution in the presence of multiple conflicts and improved resolution trajectories due to the more accurate trajectory uncertainty constraints.