Oliver D N Turnbull

Fuzzy Decision Tree Cloning of Flight Trajectory Optimisation for Rapid Path Planning

A decision tree algorithm is used to infer a set of linguistic decision rules from a set of two-dimensional obstacle avoidance trajectories optimised using mixed integer linear programming (MILP). A method to predict a discontinuous function with fuzzy decision trees is proposed and shown to make a good approximation to the optimisation behaviour with significantly reduced computational expense. Decision trees are shown to generalise to new scenarios of greater complexity than those represented in the training data and to make decisions on a time scale that would enable implementation in a real-time system. It is also demonstrated that the transparency of the rule based approach is useful in understanding the behaviour exhibited by the controller. Therefore, the decision trees are shown to have the potential to be effective online controllers for obstacle avoidance when trained on data generated by a suitable optimisation technique such as MILP

A cloned linguistic decision tree controller for real-time path planning in hostile environments

Abstract The idea of a Cloned Controller to approximate optimised control algorithms in a real-time environment is introduced. A Cloned Controller is demonstrated using Linguistic Decision Trees (LDTs) to clone a Model Predictive Controller (MPC) based on Mixed Integer Linear Programming (MILP) for Unmanned Aerial Vehicle (UAV) path planning through a hostile environment. Modifications to the LDT algorithm are proposed to account for attributes with circular domains, such as bearings, and discontinuous output functions. The cloned controller is shown to produce near optimal paths whilst significantly reducing the decision period. Further investigation shows that the cloned controller generalises to the multi-obstacle case although this can lead to situations far outside of the training dataset and consequently result in decisions with a high level of uncertainty. A modification to the algorithm to improve the performance in regions of high uncertainty is proposed and shown to further enhance generalisation. The resulting controller combines the high performance of MPC–MILP with the rapid response of an LDT while providing a degree of transparency/interpretability of the decision making.

Human Control of Air Traffic Trajectory Optimizer

Supervisory constraints are developed for a trajectory optimizer for air traffic. By choosing to apply combinations of these constraints, a human controller can exercise intuitive influence over how conflicts between aircraft are resolved. This offers a compromise between the flexibility of an automated optimizer and the insight of a human controller. Requirements, such as the sense of resolution,—i.e., which aircraft goes first or over—are encoded as constraints on a mixed-integer linear program. Examples verify that the constraints work as expected and that the computation times required are reasonable.

Wind optimal flight trajectories to minimise fuel consumption within a 3 dimensional flight network

This paper assesses the potential fuel savings benefits that can be gained from wind optimal flight trajectories. This question is posed on a 3 dimensional fixed flight network consisting of discrete waypoints which is representative of the size of Europe. The optimisation implements Dijkstras shortest path algorithm to compute the minimum fuel burn route through a network and compares this to the fuel burn for the shortest distance route. Particular effort is applied to testing the repeatability and robustness of the results. This is achieved through a sensitive analysis based on a number of identified model parameters relating to the setup of the flight network. The results of this study show fuel savings between 1.0%-10.3%, and suggest that the benefits of wind optimal flight trajectories are significant.

Does Online Search Predict Sales? Evidence from Big Data for Car Markets in Germany and the UK

We use online search data to predict car sales in the German and UK automobile industries. Search data subsume several distinct search motives, which are not separately observable. We develop a model linking search motives to observable search data and sales. The model shows that predictions of sales relying on observable search data as a proxy for prepurchase search will be biased. We show how to remove the biases and estimate the effect of pre-purchase search on sales. To assist identification of this effect, we use the introduction of scrappage subsidies for cars in 2008/2009 as a quasi-natural experiment. We also show that online search data are (i) highly persistent over time, (ii) potentially subject to permanent shocks, and (iii) correlated across products, but to different extent. We address these challenges to estimation and inference by using recent econometric methods for large N, large T panels.