Project Course in Automatic Control

A project course for master students in control and robotics, where the students during a period of 7 weeks get to work on an autonomous system.

2018 project: Model Predictive Control for a Model Car

Platform
In this project the F1/10 platform is used, which is an autonomous 1/10th scale F1 race car.

Project task
Implement Model Predictive Control for following a trajectory. This includes for example the following subparts

  • Model the system
  • Identify unknown system parameters
  • Linearize it to use Linear MPC
  • Filter data from sensors
  • Implement the optimization problem in ROS

Other than the controller, the project also includes generation of trajectories from

2017 project: Drone Search and Rescue

Motivation
A multi-agent system uses wireless communication techniques to enable collaboration between separated agents to perform tasks that a single agent could not perform.

One important application of such a system is search and rescue missions. In a search and rescue mission, several ground and aerial robots scout an area to find victims and deliver necessary supplies.

Platform
We will use the CrazyFlie2 from BitCraze.

This drone system has been used in many research projects. For example this project uses 50 CrazyFlies to demonstrate their formation flight algorithms. Here is the repository of another group that has done research with the drones.

The code from last year’s BitCraze project can be found here.

Project task
Design and implement a system with three drones; one that transports (imaginary) supplies to a rendezvous point and two that transports it further to victim sites. The time and place for the rendezvous should be determined dynamically by the system. The drones must be able to take off, land and travel the path without colliding with other drones.

Challenges

  • How to control individual drone in a smart way
  • What information should be shared?
  • Collision avoidance
  • Route planning
  • Rendezvous between drones

Interesting extentions:

  • Try to optimize path travelled, time required
  • More than one source of supplies?
  • More than one delivery drone?
  • Prioritized victims