Quantized multi-height ocuppancy grid map applied to airplanes

Abstract

The automation of cars has been widely researched and developed in the past years. Some of the technologies embedded in self-driving cars are specific for their environments, such as traffic sign recognition and lane marking detection. However, many of the technologies developed for self-driving cars are extensible to general robotic applications, directly or with some adjustments. Aiming for that opportunity, this work proposes a novel application of traditional two-dimension occupancy grid maps, widely used in self-driving cars, to the aircraft context of the taxiing operation. The main contributions are a new collision representation model and a mapping generation system. This approach handles the complex geometry of the airplane, where the wings and the body have different collision heights, the new system also introduces a more flexible collision representation model, that reduces the movement restrictions, compared to the actual system present at IARA. The proposed method also takes advantage of the two-dimensional localization, since the airplane moves basically in a 2D plane during this maneuver, which is simpler when compared to conventional three-dimensional localization and mapping systems. This work is part of an university-industry collaboration effort to unveil the technological challenges involved in automating such manoeuvres on an aircraft. The performance of the proposed method was accessed in partnership with Embraer S.A., as part of an integrated proof-ofconcept solution embedded on a Praetor 600 jet, which was able to successfully perform an autonomous taxiing over a four miles closed circuit at a private aerodrome.