Development of data-driven strategies for system analysis and control design
The main objective of the research is to find new ways in the data-modelling, parameterization-design and analysis triangle. Designing learning-based, adaptive nonlinear controls based on industry expectations, mainly for fast autonomous systems (robots, quadcopters, cars) operating in changing environmental conditions, which can be verified and validated, i.e. their reliability and quality properties (performance, robustness) can be verified is in the focus of the research.
Related to vehicle control tasks, the research topic would like to examine in what context and how it is possible to achieve some control goal starting directly from the measured data (planning) and to examine the quality of the controller already planned and implemented (performance analysis and validation). The goal is not to integrate the so-called learning algorithms, which are mostly intelligent sensors based on some kind of visual information, but mostly to study the classical sensors and the control schemes based on them. The basic question is: can we use the available measurements/data in a more intelligent (efficient) way than we have done so far, especially compared to existing procedures, e.g., adaptive, qLPV design and analysis.
Basic research in modeling, model reduction and model identification. Linear variable parameter and model predictive control methods for trajectory planning, vehicle longitudinal and transverse motion control.
Developing optimal control through machine learning. Supporting complex, computationally demanding nonlinear optimization tasks with machine learning for simpler real-time implementation.
Measurement and data processing of vehicle dynamics and traffic characteristics during a data collection campaign. Mapping public data sources. Investigating the processing capabilities of a large volume database in the cloud.
D. Fényes, B. Németh, P. Gáspár: A Novel Data-Driven Modeling and Control Design Method for Autonomous Vehicles / Energies, 14(2), 2021
A. Lelkó, B. Németh, P. Gáspár: Stability and tracking performance analysis for control systems with feed-forward neural networks / ECC'21, accepted.