Planning and reconfigurable control of a fleet of unmanned vehicles for taxi operations in airport environment

The optimization of airport operations has gained increasing interest by the aeronautical community, due to the substantial growth in the number of airport movements (landings and take-offs) experienced in the past decades all over the world. Forecasts have confirmed this trend also for the next decades. The result of the expansion of air traffic is an increasing congestion of airports, especially in taxiways and runways, leading to additional amount of fuel burnt by airplanes during taxi operations, causing additional pollution and costs for airlines. In order to reduce the impact of taxi operations, different solutions have been proposed in literature; the solution which this dissertation refers to uses autonomous electric vehicles to tow airplanes between parking lots and runways. Although several analyses have been proposed in literature, showing the feasibility and the effectiveness of this approach in reducing the environmental impact, at the beginning of the doctoral activity no solutions were proposed, on how to manage the fleet of unmanned vehicles inside the airport environment. Therefore, the research activity has focused on the development of algorithms able to provide pushback tractor (also referred as tugs) autopilots with conflict-free schedules. The main objective of the optimization algorithms is to minimize the tug energy consumption, while performing just-in-time runway operations: departing airplanes are delivered only when they can take-off and the taxi-in phase starts as soon as the aircraft clears the runway and connects to the tractor. Two models, one based on continuous time and one on discrete time evolution, were developed to simulate the taxi phases within the optimization scheme. A piecewise-linear model has also been proposed to evaluate the energy consumed by the tugs during the assigned missions. Furthermore, three optimization algorithms were developed: two hybrid versions of the particle swarm optimization and a tree search heuristic. The following functional requirements for the management algorithm were defined: the optimization model must be easily adapted to different airports with different layout (reconfigurability); the generated schedule must always be conflict-free; and the computational time required to process a time horizon of 1h must be less than 15min. In order to improve its performance, the particle swarm optimization was hybridized with a hill-climb meta-heuristic; a second hybridization was performed by means of the random variable search, an algorithm of the family of the variable neighborhood search. The neighborhood size for the random variable search was considered varying with inverse proportionality to the distance between the actual considered solution and the optimal one found so far. Finally, a tree search heuristic was developed to find the runway sequence, among all the possible sequences of take-offs and landings for a given flight schedule, which can be realized with a series of taxi trajectories that require minimum energy consumption. Given the taxi schedule generated by the aforementioned optimization algorithms a tug dispatch algorithm, assigns a vehicle to each mission. The three optimization schemes and the two mathematical models were tested on several test cases among three airports: the Turin-Caselle airport, the Milan-Malpensa airport, and the Amsterdam airport Schiphol. The cost required to perform the generated schedules using the autonomous tugs was compared to the cost required to perform the taxi using the aircraft engines. The proposed approach resulted always more convenient than the classical one.