Genetic Programming to Optimise 3D Trajectories

Abstract

Trajectory optimisation is a method of finding the optimal route connecting a start and end point. The suitability of a trajectory depends on non-intersection with any obstacles as well as predefined performance metrics. In the context of UAVs, the goal is to minimise the cost of the route, in terms of energy or time, while avoiding restricted flight zones. Artificial intelligence techniques including evolutionary computation have been applied to trajectory optimisation with various degrees of success. This thesis explores the use of genetic programming (GP) to optimise trajectories in 3D space, by encoding 3D geographic trajectories as syntax trees representing a curve. A comprehensive review of the relevant literature is presented, covering the theory and techniques of GP, as well as the principles and challenges of 3D trajectory optimisation. The main contribution of this work is the development and implementation of a novel GP algorithm using function trees to encode 3D geographical trajectories. The trajectories are validated and evaluated using a realworld dataset and multiple objectives. The results demonstrate the effectiveness of the proposed algorithm, which outperforms existing methods in terms of speed, automaticity, and robustness. Finally, insights and recommendations for future research in this area are provided, highlighting the potential for GP to be applied to other complex optimisation problems in engineering and science.