DroneHub AUDROS – Autonomous drones in CBRN applications
DroneHub AUDROS is a demonstration project for the usage of fully autonomous drones in security CBRN (Chemical, Biological, Radiological, and Nuclear) applications. The project is sponsored by the European Space Agency (ESA) and European Defense Agency (EDA). The project was preceded by a feasibility study.
A fully autonomous drone based system is that which after setting up the operating parameters, the operator no longer intervenes in drone operation. Two applications were selected for the demonstration project: applications for monitoring large-scale interventions, such as fires in industrial facilities, large-scale accidents, etc., and an automatic system for capturing enemy drones carrying potentially dangerous CBRNe cargo, spray substances, radioactive substances, or explosives. Detection of radiation using AUVs has been solved in the RADRON project at the CTU MRS group.
The project goal is the implementation of a fully functional solution in real conditions. The mounting of the sensors is modular and industrially scalable for commercial production and application.
F4F/CTU has developed a solution for capturing enemy drones.
- Military & defense – a small drone carrying a CBRN weapon represents a significant threat necessary to be dealt with,
- Prison guarding – the use of drones to traffic drugs and weapons is predicted to become a widespread issue in the coming years,
- Critical infrastructure protection – a growing number of serious incidents involving critical infrastructure, including terrorist activities and cyber threats, require new approaches to protection
An autonomous aerial intercepting system (AAIS) detects an intruder drone in a particular no-fly zone typically using radar and pre-installed cameras. The system distinguishes a drone from other objects (such as birds) and alerts an operation center. As soon as the AAIS system detects an intruder, it informs the terrestrial control station operator. The operator shall verify and confirm the intruder’s presence and decide the next steps. In case the operator decides that the intruder must be eliminated, they must give a command to start a hunter drone (RPAS-BRUS) from the docking station.
The hunter drone at the docking station is typically turned off. Once the hunter is remotely activated, its autopilot is ready for a flight in manual mode. The autopilot is able to navigate the hunter to a coordinate, typically the position of the intruder.
The intruder deactivation phase will be performed using the autonomous flight mode. The goal is to ensure a high reliability of focus and targeting of the intruder and the right timing of the attack.
The drone hunter hunts the enemy drones by catching them with a deployed net. A net gun launches and casts the net over a large space, thereby increasing the likelihood of successful intervention when compared to other solutions. The shot is triggered autonomously by an onboard computer. The trigger can also be preceded by a confirmation from the system’s operator. However, security critical applications can use the system solely in autonomous mode without any operator in order to speed up the defensive manoeuvre.
After deployment, the net remains connected to the drone hunter via a rope, preventing the caught drone from being dropped. It can then safely transport and land the target at a specific location for further investigation. Any enemy drones captured by the net are usually only slightly damaged.
Signal from the Terrestrial Control Station is transmitted over WiFi. Commands or positions are received by radio transceiver (Ethernet protocol) and forwarded to autopilot Pixhawk 4 Mavlink protocol over UART. Pixhawk is directly connected with an on-board computer, which controls the autonomous flight. If the on-board computer detects a command for starting the autonomous flight mode, it immediately takes over the flight control.
As the intruder is expected to be a moving flying object (UAV), it can happen that when the hunter reaches the given position, the intruder is already in a different position. Therefore, the terrestrial control station can update the intruder position at any time and the on-board computer of the hunter will immediately navigate to the new target position.
The operator can stop autonomous flight mode and return the drone hunter to manual flight mode at any time by transmitting the corresponding command. In that case, the hunter immediately stops the ongoing action and stays in the current position, waiting for further commands.
The Eagle.One drone hunter utilizes a depth image from a stereo camera to facilitate the onboard detection of drones and flies in their proximity. This method does not rely on using any kind of markers, which enables localization of non-cooperating drones. The output of the detection algorithm is filtered by a 3D multi-target tracking algorithm to reduce false positives, preserve temporal consistency of the detections, and to predict positions of the drones (e.g. to compensate camera and processing delays). Using neural networks, the onboard artificial intelligence is able to detect, classify, track, and predict the movement trajectory of the intruding drone in order to plan the offensive manoeuvre.
In addition to manual flight components, the following components are required for autonomous flight mode:
- Stereo camera (Intel Realsense D455)
- intruder detection, 3D localization and mapping (SLAM), computer vision
- RGB camera (1280×800 @90 fps, 1 MPx) (global shutter)
- stereo depth IR projector (1280×720 @90 fps)
- IR: range 0.2 m – 10 m (varies with lighting conditions)
- IR: Depth Diagonal Field of View over 90°
- Dot-projector allows operation in low light conditions
- integrated synchronized IMU, developed for moving cameras
- Rangefinder (Garmin LidarLite V3)
- required for autonomous flight, works as a optical distance measurement sensor
- range around ~15 m in real conditions
- resolution: 1 cm, +/- 2.5 cm at distances greater than 1 meter
- On-board computer (Intel NUC)
- handles entire autonomous flight
- compact and light all-in-one board
- integrated dual WiFi 2.4 & 5 GH
- Net Gun & Net Gun control
- an AVR/ARM microcontroller is used as a control unit
- intruder will be deactivated by being caught in a net
- the net gun itself is autonomously controlled directly by the on-board computer
- only a single net charge can be placed inside the net gun
For more details about the drone hunter, visit: http://www.eagle.one