„From Control Engineering to Intelligent Systems: Integrating AI, Learning, and Hybrid Modeling for Autonomous Machines and Processes“

Talk Abstract: Control engineering has evolved from a discipline focused on system analysis and controller design into a comprehensive science of system design. Beyond mechanical construction, sensors, actuators, control, software, and domain-specific knowledge, data analytics, machine learning, and artificial intelligence (AI) now play a pivotal role. While classical automation relies on hierarchical control loops that follow a sequential sense-analyze-calculate-act process, modern systems systema­tically integrate perception, sensor fusion, learning, and decision-making in a more dynamic and adaptive manner. Advanced optimization and control algorithms enable these systems to adapt to changing environments in real time, systematically handle nonlinear effects, and continuously operate at peak efficiency. This shift also impacts mechatronics and process control, where data-driven approaches, machine learning, and AI are increasingly shaping the evolution toward intelligent, self-optimizing systems.

The first part of the talk addresses robotic systems and presents an online model-predictive planner for Cartesian reference paths in the end-effector’s pose. This planner operates robustly in dynamic environments and under varying task constraints, providing an effective interface between low-level control and high-level reasoning, including integration with large language models. Applications to large-scale autonomous machines, in particular automated pallet loading by outdoor forklifts and log handling by truck-mounted cranes, demonstrate how the tight coupling of perception, planning, and control enables high levels of autonomy in both industrial and natural settings.

The second part focuses on hybrid modeling approaches that combine first-principles and data-driven methods in process control. By fusing multi-modal sensor data with physics-based and machine learning models, it becomes possible to dynamically estimate unmeasurable product properties during manufacturing. This capability forms the foundation for advanced control concepts that adaptively correct deviations in real time and compensate for variations in raw materials or upstream disturbances. Examples from the steel industry illustrate how such approaches can significantly improve process stability, product quality, and energy efficiency.