Aerial vehicles have been designed to meet all the requirements of Data Acquisition Service module and are custom made by a company Kopterworx, which is one of the industry leaders in the field (Figure 1). The frame consists of four carbon arms and two legs with skis. The vehicle is equipped with Flame 60A 12S ESCs, which are driving T-motor U8 Lite Kv150 12S motors with 22 inch carbon propellers. Vehicle autopilot is a Pixhawk 2.0 with Ardupilot software which uses data from the navigation sensors which include IMU, barometric sensor, GPS receiver, magnetometers and gyroscopes. Maximal take-off weight of the vehicle is 12 kg with 2 kg of payload. Vehicle is equipped with Intel NUC i7/16GB computer running Ubuntu 18.04 LTS and ROS Melodic, which is processing inputs from ZED Mini stereo camera, Smartek visual camera and Velodyne VLP-16 Puck Lite 3D lidar (Figure 2). Image processing for UAV is performed using convolutional neural networks running on Intel Neural Compute Stick 2. All components of the aerial vehicle are powered by two LiPo 12S 14000 mAh batteries, giving the vehicle up to 30 minutes of flight time.
For the development and evaluation of algorithms, we are using desktop and laptop workstations running Ubuntu 18.04 LTS and ROS Melodic with Gazebo 9 simulator. To minimize transition time from simulator to real vehicles we are employing realistic vehicle and sensor models and software-in-the-loop simulation.
Figure 1. Kopterworx UAV 
Figure 2. Initial 3D model based on non-processed LIDAR data acquired by UAV
