Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Model Predictive Control approach for whole body motion planning of ANYmal on rough terrain
Our goal is to enable our quadrupedal robot, ANYmal, to navigate in challenging terrain such as rubble, stairs etc. This project will allow ANYmal to safely overcome obstacles of significant size.
Keywords: Robotics, MPC, ANYmal, rough terrain,
In this project, we will focus on continuously adapting the swing leg motion and main body posture of the robot as well as the ground contact force distribution to navigate over cluttered terrain safely. To this end, we will use a Model Predictive Control (MPC) approach to plan the whole body motion of the robot. While motion planning in an MPC fashion for legged systems has made significant progress in last years, its application was limited to flat terrain. In this project, we will extend our current MPC implementation for ANYmal to exploit terrain information in order to adjust swing leg trajectories and body motion.
In this project, we will focus on continuously adapting the swing leg motion and main body posture of the robot as well as the ground contact force distribution to navigate over cluttered terrain safely. To this end, we will use a Model Predictive Control (MPC) approach to plan the whole body motion of the robot. While motion planning in an MPC fashion for legged systems has made significant progress in last years, its application was limited to flat terrain. In this project, we will extend our current MPC implementation for ANYmal to exploit terrain information in order to adjust swing leg trajectories and body motion.
After a literature study about existing tools and methodologies, the student will start implementing appropriate algorithms in an existing MPC toolbox and simulation environment (ROS, Gazebo). After a successful implementation, the proposed algorithms will be tested on the legged robot, ANYmal, in different rough terrain scenarios.
After a literature study about existing tools and methodologies, the student will start implementing appropriate algorithms in an existing MPC toolbox and simulation environment (ROS, Gazebo). After a successful implementation, the proposed algorithms will be tested on the legged robot, ANYmal, in different rough terrain scenarios.
- knowledge of advanced control methods (e.g. MPC, Dynamic Programming and Optimal Control etc.).
- knowledge in programming (C++).
- highly motivated and excellent students.
- knowledge in ROS and Gazebo are advantageous.
- knowledge of advanced control methods (e.g. MPC, Dynamic Programming and Optimal Control etc.). - knowledge in programming (C++). - highly motivated and excellent students. - knowledge in ROS and Gazebo are advantageous.