
Can AI Control an Unstable Robot? From Classical Control to Reinforcement Learning
Webinar Details
Friday, 6 November 2026 10:00-10:45
Artificial intelligence and data-driven methods are becoming an increasingly important research direction in control engineering. But how well can these approaches actually control a dynamic system compared with established classical control methods?
In this webinar, we will use the Quanser QUBE and its unstable dynamics to explore the transition from classical control toward AI-based control. Starting with a conventional controller for stabilizing the system, we will then look at how a data-driven approach such as reinforcement learning can learn a control policy through interaction with the system.
The webinar will focus on understanding the differences between model-based classical control and data-driven control, and on evaluating how an AI-based controller performs when applied to the same control problem. We will also discuss the role of simulation, training data and real-time experimentation when developing AI-based control systems.
What you’ll learn
- Classical vs. AI-based approaches to controlling an unstable system
- How reinforcement learning can learn a control policy from system interaction
- How AI-based controllers perform compared with classical control methods
- Key challenges in transferring and validating data-driven control on real hardware
Presenter
Harshith Maregowda
Harshith Maregowda is a Mechatronics Engineer specializing in control systems, robotics, reinforcement learning, and real-time applications. He holds an M.Sc. in Mechatronics from Hamburg University of Technology (TUHH), Germany.
At made|for|science, his work focuses on developing technical solutions based on Quanser engineering platforms and interactive laboratory systems. He works at the intersection of control theory, simulation, and real-time hardware, translating theoretical concepts into practical solutions for engineering education and research.







