Project & Assignments

Course Semester Project Topics [in teams of 3-5 students]

Thematic assignments [1-2 students per assignment, each student picks 2 assignments in total]

Assignment 1. Robot Learning: Time-optimal Position Control of Multirotor Vehicles

• Goal: Learn position control policies for multirotor aerial vehicles
• Purpose: This exercise aims to expose students in the procedures involved to learn control policies.

Tasks:

• Set-up the Aerial Gym Simulator
  • https://github.com/ntnu-arl/aerial_gym_simulator
• Set-up rl-games which will be used to train the policy
  • https://pypi.org/project/rl-games/1.0.2/
• Train a position control policy for a quadrotor vehicle or a fully-actuated multirotor as per the provided examples
  • https://ntnu-arl.github.io/aerial_gym_simulator/6_rl_training/#train-your-own-position-control-policies
  • Modify the reward function to achieve time-optimized position control
  • Propose further improvements for time-optimal control
• Provide a short (4-pages max) report on the achieved performance and outline avenues to improve the training quality. Further comment on identified sources of possible sim2real gap.

Note: If you do not have access to a computer with a powerful GPU and overall specifications as needed by Aerial Gym (provide a short definition here) contact the Autonomous Robots Lab to get remote access to one of our systems. In this case, the first two points are not applicable.

Assignment 2. State Estimation: EKF, IEKF Implementation
This assignment involves creating iPython notebooks for:

• Implementation of the Extended Kalman Filter for multiple input, multiple output nonlinear systems
• Application of the EKF for 3D free-floating rigid body system (e.g., equations of motion of a quadrotor)
• Implementation of the Iterated Extended Kalman Filter for multiple input, multiple output nonlinear systems
• Application of the IEKF for the same problem as the EKF

Upload your files in blackboard and ideally share your git username so that you can be invited to push into a repository for the course.  

Assignment 3. State Estimation: Lie Theory presentation
Study and present (20min presentation) the following paper:

• Sola, J., Deray, J. and Atchuthan, D., 2018. A micro Lie theory for state estimation in robotics. arXiv preprint arXiv:1812.01537. [link]

NB! The course material is currently being added!

---