Motion Planning

Multiplex Twinstar in autonomous flight

 


 

Collision-free NMPC Motion Planning

Nonlinear Model Predictive Control (NMPC) is a control methodology that is able to consider input and state constraints explicitly. For that reason, obstacles (constraints on position of the vehicle) can be efficiently included into the motion planning problem while considering other constraints like maximum velocity e.g. at the same time. The video on the left demonstrates a flight of an octocopter in a part of the Grand Canyon to a specific waypoint avoiding the challenging terrain during the flight.

The algorithm was demonstrated on an octocopter of the Institute, as well. An artificial obstacle was used to demonstrate collision avoidance. The whole flight including take-off and landing was done without intervention of the safety pilot. The video below may give an impression of the outdoor demonstration of the NMPC motion planning.

  • Seiferth, Christoph, Alexander Joos, Michael Frangenberg und Walter Fichter. “Predictive Motion Planning with Pipelined Feature-Based Obstacle Avoidance,” Journal of Guidance, Control and Dynamics, vol. 39, no. 4, pp. 970–978, 2015.
  • A. Joos, C. Seiferth, L. Schmitt, and W. Fichter, “Parameters for Nonlinear Model Predictive Control in UAV Path Planning Applications,” AIAA Journal of Guidance, Control, and Dynamics, accepted for publication, 2016.

 


 

Finite-Receding Horizon Incremental Sampling Tree

Explanation will follow soon.

  • Gros, M., Schoettl, A., Fichter, W.: Spline and OBB-based Path-Planning for Small UAVs with the Finite Receding-Horizon Incremental-Sampling Tree Algorithm. AIAA 2013-4788, AIAA Guidance, Navigation, and Control Conference, 19-22 August 2013, Boston, Massachusetts, USA.

 


 

Automatic Landing using Visual Servoing

Marker Setup for vision aided automatic Landing

Automatic landing is an important functional element for the usability of future micro aerial vehicles (MAV). This method makes an automatic landing for MAV with inertial navigation system, pressure sensor and camera possible by placing three visual markers on the ground depending on current wind situation and local obstacle placement.

The video below shows an automatic approach and landing.

  • Trittler, M., Rothermel, T., Fichter, W.: Visual Servoing Based Landing Approach Controller for Fixed-Wing MAVs, 19th IFAC Symposium on Automatic Control in Aerospace, Sept. 2-6, 2013, Würzburg, Germany.

 


 

Landing Spot Detection

Detection of feasible Landing Spots from Elevation Map

When flying into an unknown area, an autonomous vehicle – for example a space probe landing on a foreign planet – needs some way to identify a suitable landing spot. At the institute an algorithm was developed that analyses a given digital elevation map (DEM) for inclination and roughness. If a sector is flat and smooth enough, it is marked as landable. The center of the biggest contiguously landable area is selected as the landing spot.

 

@2015 iFR - Flight Mechanics and Controls Lab

Address

Flight Mechanics and Controls Lab
Pfaffenwaldring 27
70569 Stuttgart
Germany

Contact

Phone:   +49 711 685-67061

Fax:        +49 711 685-66670

Mail:       ifr@ifr.uni-stuttgart.de