Human-Centered Robotics colloquium - Oran Liu, University of Groningen

Title: Accelerating Human-in-the-Loop Optimization Through Real-Time User Preference Feedback

Abstract:

Human-in-the-loop optimization (HILO) has emerged as a powerful framework for personalizing control strategies in wearable robotics and assistive devices. Traditional HILO approaches often rely on metabolic measurements or performance-based metrics, which require long sampling periods and may reflect user's preference. These constraints limit the speed and practicality of real-time optimization during everyday use. In this talk, I present a preference-based HILO framework that incorporates direct user feedback as the cost function, enabling fast and user-centered optimization of device control parameters. By leveraging simple preference queries rather than explicit performance measurements, the method can converge rapidly while capturing subjective factors such as comfort, stability, and perceived effort—elements that are critical for long-term device acceptance. I will introduce the underlying optimization algorithm, evidence demonstrating rapid convergence, and insights into how user preference correlates with biomechanical and energetic outcomes. This approach offers a practical pathway toward real-time personalization and more intuitive, human-aligned assistive technologies.