Mainblades — Since 2015 Mainblades Inspections (MBI) focused its business on the inspection of aircrafts. MBI aims to employ robots in this operation to reduce the aircraft’s latency time and the danger for the human operators. The robots that better fit these requirements are drones.
Drone — The usage of drones for inspection of large areas and/or objects increased exponentially in the last years. The enormous progress made on this technology made them a valid solution for inspections. The drones (also known as Micro Aerial Vehicles, MAVs) have three main characteristics that made them really competitive compared to ground robots:
Agility. Drones have the possibility to move in a 3D environment with 6DoF without any bond to the ground.
Versatility. MAVs can be used for indoor and outdoor solutions. If combined with sophisticated control algorithms these robots can fly also with adverse weather conditions, such as rain, wind etc.
Speed. The movement generated by the motor of the propellers can give a speed up to 150km/h.
One of the main drawback of this robots it’s the localization in indoor environments. In fact, when the robot is located outdoor the GPS sensor can be used to estimate the position with relatively good accuracy. However, for indoor applications, the GPS signal is not reliable anymore. The execution of inspection are often executed in large hangars located in the airport, and indeed the GPS cannot be used for the pose estimation of the robot.
LiDAR — To overcome this lack, other sensors can be used, such as LiDAR. The laser technology is already widely used in robotics for mapping and autonomous navigation, especially in the autonomous car’s field. The robot can use the information obtained by these sensors to recognize static features in the environment and determine its position with respect to them. MBI decided to solve the localization problem installing a 2D Laser Range Finder (LRF) on the robot. The laser is facing downward and it spins around the z-axis to achieve a 360 degrees field of view. The spinning movement is actuated by a motor that connects the laser to the drone.
Tracking and Mapping (TAM)— The information received by the LRF are processed by sophisticated algorithms that integrate the data provided by the Inertial Measurement Unit on the drone with the laser data to solve the problem of the localization. This process is named Tracking and Mapping and its composed of three main steps:
Registration. The scan provided by the laser is translated into a point cloud in the environment.
Odometry. The state estimation of the robot is executed using a scan matching technique, which is able to find correspondences between two consecutive scans and estimate a correction in the positioning of the robot.
Mapping. The scan is indeed re-projected in the space using the odometry computed in the previous step and it is also realigned to the previous scans received to produce a map in output.
The results of the algorithms are visible on the first video of TAM published on our YouTube Channel (follow us on our YouTube page). The data-set has been recorded in the hangar of KLM, using the first version of the drone for aircraft inspection developed by MBI.
