B-Spline Based Motion Planning and Model Predictive Control

Employer
KU Leuven
Location
Europe
Posted
March 14 2017
Position Type
Full Time
Organization Type
Academia

For the Production Engineering, Machine Design and Automation (PMA) Section we are looking for a young, motivated and skilled PhD researcher with a strong background in numerical optimization, systems theory and control. You will be embedded in the MECO research team of the KU Leuven Department of Mechanical Engineering. The MECO research team focusses on the identification, analysis and control of mechatronic systems such as machine tools, active suspensions, robots... Herein theoretical developments are combined with experimental validations on lab-scale as well as industrial setups.

In motion planning one seeks for the fastest, most energy efficient… trajectory to move a motion system from its current position to its destination, while accounting for the system's kinematic and dynamic constraints and avoiding collisions with all obstacles in the environment. It plays a vital role in the control of autonomous guided vehicles, CNC machine tools, serial robots… As motion planning is often performed on line, in a model predictive control fashion, solving the resulting optimization problems efficiently and reliably is of the utmost importance. Recently, we have developed an effective motion planning method based on B-splines. The motion trajectory is parameterized as a polynomial spline and the properties of the B-spline basis functions are exploited to efficiently enforce constraints over the considered time horizon. The method is implemented in the open source Python toolbox OMGtools (https://github.com/meco-group/omg-tools). In this research project you will extend this motion planning approach to a broad range of systems, and in collaboration with experts in numerical optimization, you will supply it with tailored optimization routines. In addition, you will contribute to the underlying B-spline based optimization approaches and explore additional uses of these approaches in model predictive control.



This job comes from a partnership with Science Magazine and Euraxess