Overview

Path planning and trajectory optimization is a fundamental task for operating unmanned aircraft systems (UAS). The aim is to design trajectories and optimization-based feedback controllers, which are able to obey constraints and are efficient with respect to a given objective function. Herein, constraints usually result from technical or physical limitations (e.g. dynamic pressure, load factor, control bounds) or are imposed to avoid collisions or no-fly zones. Typical objective functions include, e.g., energy consumption, time efficiency, noise emission, or detection probability. We apply numerical optimal control techniques and model-predictive control schemes. More specifically we develop algorithms and software packages for the following problem classes: 

  • real-time optimization methods and sensitivity analysis for path planning
  • linear and nonlinear model-predictive control (MPC) for path following and path planning
  • imitation learning model-predictive control (IL-MPC)
  • reinforcement learning (RL) for path planning and collision avoidance
  • dynamic programming (DP) for path planning and dynamic scheduling
  • dynamic scheduling and dynamic vehicle routing for the coordination of multi-agent systems


The methods are applied to the following scenarios (collected examples are provided below):

  • reconnaissance maneuvers
  • co-operative UGV and UAS maneuvers
  • coordination of multi-agent systems
  • obstacle avoidance for quadrocopters and helicopters

 

We use the robot operating system ROS2 and the Unreal Engine for visualization.

Landing maneuver of UAV on UGV in ROS2 realtime environment

The video shows a landing maneuver of a UAV on a moving UGV using an online NMPC controller for the UAV. The communication is organized in ROS2 in realtime. The UGV communicates its predicted path to the UAV, which reacts on the motion of the UGV. The visualization is done by the unreal engine with the aid of Omid Moslehirad.

Cooperative landing maneuver of UAV on UGV

The video shows a landing maneuver of a UAV on a moving UGV. Both agents are controlled simultaneously by model-predictive control (MPC) in a cooperative way. The approach is simulated in the 3D visualization environment ROCS and the unreal engine with the aid of Omid Moslehirad.

UAV Mission

The video illustrates an observation mission of a UAV with cloud avoidance. The flight path is generated by a model-predictive control technique. The approach is simulated in the 3D visualization environment ROCS and the unreal engine with the aid of Omid Moslehirad.

Flight test with mpc-generated tunnel in the sky

The video shows results from a flight test. Model-predictive control is used to generate collision-free flight paths in realtime. The desired flight paths are visualized to the pilot using a tunnel in the sky, which the pilot aims to follow. The method is capable to avoid obstacles, which can be created from a ground station using a data link.

NMPC Flight Path Optimization of a UAV

The animation shows the flight of a UAV within a flight corridor. The flight path was computed by a nonlinear model-predictive control method for a point-mass model of a quadrocopter-like UAV.

Dynamic routing problem with profits

The video shows a dynamic routing problem with profits. This is a bi-level optimization problem coupling a routing problem with optimal control problems for path planning of multiple agents. The aim is to maximize profits within a given time frame. The approach is simulated in the 3D visualization environment ROCS and the unreal engine with the aid of Omid Moslehirad.

Dynamic multi-agent travelling salesman problem with time-varying no-fly zones

The video shows a dynamic travelling salesman problem for a multi-agent system with time varying no-fly zones. The dynamic travelling salesman problem is a bi-level optimization problem coupling a multi-agent traveling salesman problem with optimal control problems for path planning.The approach is simulated in the 3D visualization environment ROCS and the unreal engine with the aid of Omid Moslehirad.

Dynamic scheduling of drones

The video shows the dynamic scheduling of drones with target assignment and collision avoidance. Dynamic scheduling combines scheduling with path planning and uses a bi-level optimization formulation. The results have been obtained by the Engineering Mathematics Group at the University of the Bundeswehr Munich.