Publications [#344897] of Silvia Ferrari
Papers Published
- Liu, C; Ferrari, S, Vision-guided planning and control for autonomous taxiing via convolutional neural networks,
Aiaa Scitech 2019 Forum
(January, 2019), ISBN 9781624105784 [doi]
(last updated on 2021/09/05)Abstract:
This paper presents a new approach for autonomous taxiing that can potentially improve the efficiency and safety of ground operations in commercial or military airports. Research on airport automation so far has focused primarily on scheduling and coordination of piloted aircraft for improved overall efficiency and lower operational costs. This paper develops a novel vision-guided path planning and control approach that could potentially lead to unmanned taxiing and takeoff. Autonomous taxing is a challenging problem because on-board perception algorithms must be capable of translating verbal commands provided by the Air Traffic Control (ATC) tower and, at the same time, react to a wide range of possible runway incursions, taxiway and runway conditions, and ground crew behaviors, in order to operate safely in complex airport environments without human intervention. In this paper, the autonomous taxiing problem is formulated as a hybrid planning and control problem that can be solved by an approach based on vision-based perception, obstacle avoidance, and feedback control theory. By harnessing convolutional neural networks (CNNs) for computer vision, the information obtained by onboard cameras about surrounding environments can be integrated with prior information – such as airport maps – and ATC commands to compute motion plans, while simultaneously detecting and adapting to dangerous situations such as runway incursions. Simulation results obtained using the photo-realistic physics-based Unreal EngineT M simulation tool show that the proposed approach can be potentially used some day to automate airport ground operations, even in crowded environments populated with ground crew, vehicles, and other aircraft.