Regularized GLISp for sensor-guided human-in-the-loop optimization

2025-11-06

AI Summary

Existing preference-based optimization methods (e.g., GLISp) treat the system as a black box, disregarding observable sensor measurements—leading to slow convergence and suboptimal solutions. To address this, we propose a regularized GLISp framework that incorporates measurable physical quantities: a gray-box model is constructed by embedding sensor data into the preference learning process via physics-informed hypothesis functions and a least-squares regularization term. This constitutes the first explicit integration of objective sensor information within a preference optimization framework, effectively balancing the flexibility of human or expert preference feedback with the guidance of physical priors. Evaluated on analytical benchmarks and a real-world vehicle suspension tuning task, the proposed method achieves significantly faster convergence and superior final performance compared to standard GLISp.

Technology Category

• Knowledge Representation and Reasoning: Preferences
• Search and Optimization: Learning to Search
• Machine Learning: Learning Preferences or Rankings

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs

Abstract

Human-in-the-loop calibration is often addressed via preference-based optimization, where algorithms learn from pairwise comparisons rather than explicit cost evaluations. While effective, methods such as Preferential Bayesian Optimization or Global optimization based on active preference learning with radial basis functions (GLISp) treat the system as a black box and ignore informative sensor measurements. In this work, we introduce a sensor-guided regularized extension of GLISp that integrates measurable descriptors into the preference-learning loop through a physics-informed hypothesis function and a least-squares regularization term. This injects grey-box structure, combining subjective feedback with quantitative sensor information while preserving the flexibility of preference-based search. Numerical evaluations on an analytical benchmark and on a human-in-the-loop vehicle suspension tuning task show faster convergence and superior final solutions compared to baseline GLISp.

Problem

• Integrating sensor measurements into preference-based optimization loops
• Combining subjective human feedback with quantitative sensor data
• Accelerating convergence in human-in-the-loop calibration tasks

Innovation

• Integrates sensor data into preference learning loop
• Uses physics-informed hypothesis function for guidance
• Adds least-squares regularization for grey-box structure