Nonlinear MPC for Quadrotor Fault-Tolerant Control (RAL 2022)

The mechanical simplicity, hover capabilities, and high agility of quadrotors lead to a fast adaption in the industry for inspection, exploration, and urban aerial mobility. On the other hand, the unstable and underactuated dynamics of quadrotors render them highly susceptible to system faults, especially rotor failures. In this work, we propose a fault-tolerant controller using the nonlinear model predictive control (NMPC) to stabilize and control a quadrotor subjected to the complete failure of a single rotor. Differently from existing works that either rely on linear assumptions or resort to cascaded structures neglecting input constraints in the outer-loop, our method leverages full nonlinear dynamics of the damaged quadrotor and considers the thrust constraint of each rotor. Hence, this method can seamlessly transition from nominal to rotor failure flights, and effectively perform upset recovery from extreme initial conditions. Extensive simulations and real-world experiments are conducted for validation, which demonstrates that the proposed NMPC method can effectively recover the damaged quadrotor even if the failure occurs during aggressive maneuvers, such as flipping and tracking agile trajectories. If you are interested to know how to restabilize a quadrotor subject to rotor failure with onboard cameras, check our previous work: • Autonomous Quadrotor F...
Reference
F. Nan, S. Sun, P. Foehn, D. Scaramuzza
Nonlinear MPC for Quadrotor Fault-Tolerant Control
Robotics and Automation Letters (RAL), 2022
Paper: rpg.ifi.uzh.ch...
More on our research in Agile Drone Flight: rpg.ifi.uzh.ch/...
More on our research in Autonomous Drone Racing: rpg.ifi.uzh.ch/...
Affiliations:
F. Nan, S. Sun, and D. Scaramuzza are with the Robotics and Perception Group, Dep. of Informatics, University of Zurich, and Dep. of Neuroinformatics, University of Zurich and ETH Zurich, Switzerland
rpg.ifi.uzh.ch/

[@LeRainbow]

Incredible! This is how we make technology safe. By teaching it how to deal with the non-boundary solutions when the so to say "fecal matter hits the rotary impeller". Can't wait what else you come up with. Also: thank you very much for PUBLICLY including the control schemata to recreate this. I love research when it is done well and free, you are awesome!

[@munirjojo-verge4295]

😂

[@hopsenrobsen]

The same principle was already demonstrated back in 2013 by Raffaello D'Andrea (search for "the athletic powers of quadcopters"). Nevertheless, congrats for achieving this! The real challenge is to make it work without computer vision, in an uncontrolled enviroment, that is what I would love to see. All the best!

[@anan4866]

Actually, the real challenge is to make it work during aggressive flights. As the work from Raff used a linear controller that would fail if the initial condition is not near-hovering.

[@hopsenrobsen]

@@anan4866 Good point but, for drones out in the wild there is no computer-vision, right? Do you think it will be achievable to realize this typ of failsafe using onboard sensor data only?

[@sihaosun9491]

@@hopsenrobsen Please check this video th-cam.com/video/Ww8u0KH7Ugs/w-d-xo.html

[@davidescaramuzza]

@@hopsenrobsen We already did this with onboard cameras. The link is in the video description: " If you are interested to know how to restabilize a quadrotor subject to rotor failure with onboard cameras, check our previous work: th-cam.com/video/Ww8u0KH7Ugs/w-d-xo.html"
Now combine these two works and you have a system that uses onboard cameras for aggressive flight of quadrotors with rotor failure.

[@yingfuxu8466]

Happy to see it flies in a trajectory using 3 motors haha

Thank you for sharing.

[@SixteenVoice]

Thank you!

[@arunbali7480]

Very impressive lecture
Sir I have a question.
In practice, faults are often unpredictable, and its specific form is difficult to model. Is it too simple to use efficiency and bias to describe faults?

[@snakehaihai]

dji M300 had similar one motor failure test outdoor at late 2019. What your improvement when compare with their performance?

[@howrobotics2052]

th-cam.com/video/4IEvSzVRrsQ/w-d-xo.html

[@sihaosun9491]

Well... I gave a talk to DJI and had some discussions with them regarding fault-tolerant control several months before they release this video. So based on my understanding of their controller, I don't think DJI's solution can recover it if the drone is almost upside-down, which might happen after a collision with other drones etc. Motor failure is very unlikely to happen without a collision after all.

[@snakehaihai]

@@sihaosun9491 Well thanks for you answer. Did you include this in your paper?

[@anan4866]

​@@snakehaihai Unfortunately we didn't, as their method on a commercial product is not an open research. However, we did analysed existing methods from published literatures. Existing methods either suffers from suboptimal performance (eg doesn't recover from arbitrary attitude) or have never been implemented in real-world experiments.