Justified Text Example

Planning and Decision Making
for Autonomous Robots


Ivan Alberico     Nicola Loi    

Institute for Dynamic Systems and Control, ETHz


Task

Control a spacecraft to safely reach the goal region, moving through an environment full of static and dynamic obstacles. The spacecraft has a left and a right thruster at the back that can be activated to move forward or backward. The simulator provides observations about the state of the spacecraft and of the moving obstacles which correspond to the real state of the simulation. Optimize the distance traveled, the final time, and the actuation effort.



Pipeline


  • To complete the task, at the beginning of the simulation an Informed RRT* algorithm is applied to the static environment only, to obtain the global path the spacecraft has to follow to reach the goal region, given a safe margin from obstacles. The global path found is then smoothed.
  • A MPC is exploited to calculate an optimal local trajectory to track the smoothed global path while avoiding the dynamic obstacles. The inputs of the *n*-steps horizon MPC are the pose and velocity of the spacecraft, the next *n* points of the smoothed path, and the next *n* future positions of the dynamic obstacles (predicted through their current pose/velocity and dynamics). The spacecraft is constrained to remain inside the environment boundaries, and to not exceed the speed of 50 km/h.

Video examples

Example 1



Example 2



Example 3