Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Control and State Estimation for Automated In-Flight Payload Recovery with a Fixed-Wing UAV
This project will focus on state estimation, the respective sensing as well as control of a fixed-wing unmanned aerial vehicle to perform in-flight pick-ups of small payloads.
The fixed-wing team of the Autonomous Systems Lab (ASL) is actively involved with small fixed-wing unmanned aerial vehicle (UAV) research. The endurance and high speed of fixed-wing flight is promising for long-range delivery of, e.g., parcels or the deployment of sensors for remote-sensing. However, getting the payload on the aircraft in the first place usually requires the aircraft to land, be loaded (typically by an external system), and then launched again by operators on the ground. In cases where the payloads need to be repeatedly picked and placed (e.g. sensors) or when they are located in regions inaccessible to ground operators, it would be desirable to automate this (re)loading process. To this end, we envision equipping our fixed-wing UAVs with the necessary hardware and algorithms to approach and pick small payloads in-flight as do, for example, birds of prey.
In a recent project, students developed and flight-tested a combination of catching mechanism and payload which allows for repeated drop- and recovery [Simon Heinzmann & Timon Wehmann, BSc. Mech. Eng. ETHZ]. This work demonstrated the general feasibility of in-flight ‘pick-and-place’ operation with manually-piloted fixed-wing UAVs, cf. attached pictures. One of the big remaining challenges comprises precise control of the aircraft’s approach trajectory and, consequently, accurate state-estimation/localization with respect to the payload and terrain. The goal of this project is to select and integrate the sensors required for the said state-estimation and augment the flight-control system to perform pick-ups with high reliability. This can include methods such as visual-servoing or more complex planning- and tracking of feasible approach trajectories. The system will be flight-tested on one of our UAVs equipped with the previously-developed pick-up interface and, ideally, outperform the 50% pick-up rate achieved by an experienced human remote-controlling (RC) pilot.
The fixed-wing team of the Autonomous Systems Lab (ASL) is actively involved with small fixed-wing unmanned aerial vehicle (UAV) research. The endurance and high speed of fixed-wing flight is promising for long-range delivery of, e.g., parcels or the deployment of sensors for remote-sensing. However, getting the payload on the aircraft in the first place usually requires the aircraft to land, be loaded (typically by an external system), and then launched again by operators on the ground. In cases where the payloads need to be repeatedly picked and placed (e.g. sensors) or when they are located in regions inaccessible to ground operators, it would be desirable to automate this (re)loading process. To this end, we envision equipping our fixed-wing UAVs with the necessary hardware and algorithms to approach and pick small payloads in-flight as do, for example, birds of prey. In a recent project, students developed and flight-tested a combination of catching mechanism and payload which allows for repeated drop- and recovery [Simon Heinzmann & Timon Wehmann, BSc. Mech. Eng. ETHZ]. This work demonstrated the general feasibility of in-flight ‘pick-and-place’ operation with manually-piloted fixed-wing UAVs, cf. attached pictures. One of the big remaining challenges comprises precise control of the aircraft’s approach trajectory and, consequently, accurate state-estimation/localization with respect to the payload and terrain. The goal of this project is to select and integrate the sensors required for the said state-estimation and augment the flight-control system to perform pick-ups with high reliability. This can include methods such as visual-servoing or more complex planning- and tracking of feasible approach trajectories. The system will be flight-tested on one of our UAVs equipped with the previously-developed pick-up interface and, ideally, outperform the 50% pick-up rate achieved by an experienced human remote-controlling (RC) pilot.
1. Identify requirements of state-estimation for payload pick-up (for simplicity, we constrain the problem to environments with flat ground and no proximal obstacles)
2. Identify and integrate suitable sensors to enable state-estimation (e.g. monocular vision, GNSS, INS)
3. Develop and implement control strategies to fly precise pick-up approaches.
4. Demonstrate the developed system in real-world flight-tests and pick-up payloads
1. Identify requirements of state-estimation for payload pick-up (for simplicity, we constrain the problem to environments with flat ground and no proximal obstacles) 2. Identify and integrate suitable sensors to enable state-estimation (e.g. monocular vision, GNSS, INS) 3. Develop and implement control strategies to fly precise pick-up approaches. 4. Demonstrate the developed system in real-world flight-tests and pick-up payloads
- Strong interest and background in control and state-estimation
- Familiar with C++, ideally ROS, and the PX4 flight-stack.
- Experience and hands-on attitude with hardware integration, e.g. with Raspberry Pi
- Understanding of the basic fixed-wing flight mechanics
- Ideally, RC model airplane piloting experience as this facilitates flight-testing
Note: This project is primarily intended as a Master Thesis. Optionally, a part of this project can be conducted in a concurrent Semester Thesis which focuses on the selection- and integration of the sensor system on the UAV.
- Strong interest and background in control and state-estimation - Familiar with C++, ideally ROS, and the PX4 flight-stack. - Experience and hands-on attitude with hardware integration, e.g. with Raspberry Pi - Understanding of the basic fixed-wing flight mechanics - Ideally, RC model airplane piloting experience as this facilitates flight-testing
Note: This project is primarily intended as a Master Thesis. Optionally, a part of this project can be conducted in a concurrent Semester Thesis which focuses on the selection- and integration of the sensor system on the UAV.
David Rohr: david.rohr@mavt.ethz.ch
Thomas Stastny: thomas.stastny@mavt.ethz.ch
Nicholas Lawrance: nicholas.lawrance@mavt.ethz.ch
David Rohr: david.rohr@mavt.ethz.ch Thomas Stastny: thomas.stastny@mavt.ethz.ch Nicholas Lawrance: nicholas.lawrance@mavt.ethz.ch