Workshop on optimal control problems and model predictive control for autonomous systems

Abstract

In this workshop, participants will engage in hands-on exploration of optimal control problems (OCPs) applied to motion planning and model predictive control (MPC) in autonomous robotic systems. By engaging with cutting-edge tools and techniques, participants will develop the skills necessary to navigate complex environments, optimize trajectory paths, and execute tasks with precision and efficiency in robotic systems. To streamline the guided exercises, the workshop makes use of the free and open-source Rockit and Impact software frameworks developed by the MECO Research Team at KU Leuven and built on top of the numerical optimization framework CasADi, designed for efficient nonlinear programming. Exercises will be mainly in Python, and Docker images containing a development and simulation environment will be provided. Attendees can later adopt the presented open-source software frameworks in their research. While foundational concepts of OCPs will be introduced, the course focuses on learning-by-doing. The course prioritizes practical know-how, enabling participants to directly apply Rockit and Impact to tackle real-world robotic challenges. The attendees will learn to formulate and solve OCPs, gaining valuable experience in implementing trajectory optimization algorithms and MPC strategies. Moreover, participants will learn how to swiftly deploy OCPs and MPCs in C, Python and ROS 2. This workshop is organized by members of the MECO Research Team of KU Leuven, Belgium. The MECO Research Team focusses on modeling, estimation, identification, analysis and optimal control of motion and motion systems such as mechatronic systems or machine tools. It combines theoretical contributions (development of design methodologies) with experimental knowhow (implementation and experimental validation on lab-scale as well as industrial setups). The theoretical research benefits from the group’s expertise on numerical optimization, especially convex optimization.