Abstract:

Koopman operator theory has emerged as a powerful tool for modelling and controlling nonlinear dynamical systems, offering a way to transform complex, nonlinear behaviours into linear representations. In robotics, this theory holds significant promise for enhancing control and prediction in tasks such as manipulation, where traditional methods struggle with the complexities of nonlinearity. By applying the Koopman operator, robotic systems can model the dynamics of interactions between robots and objects in a higher-dimensional space, enabling the use of linear techniques for more efficient control and prediction. This Review explores the application of Koopman operators in robotics, particularly in dexterous manipulation, where the challenge lies in learning and predicting object states that are often difficult to capture. The review discusses the advantages of Koopman operators in simplifying system dynamics, their integration with machine learning models for real-time tasks, and their role in making robotic manipulation more data-efficient.

Speaker Bio:

Mr. Mohamed MohamedAhmed is currently pursuing his MSc. in the Department of Control and Instrumentation Engineering (CIE) at King Fahd University of Petroleum and Minerals (KFUPM), Saudi Arabia. He received his B.S degree in Electrical and Electronics Engineering from the University of Khartoum, Sudan. He worked as a Systems Engineer at MT Industrial Supplies and Services, Sudan from 2021 to 2023. His research interest includes Data-Driven Modelling and Control of Complex Systems.