Plug-and-Play Physics-informed Learning using Uncertainty Quantified Port-Hamiltonian Models

2025-04-24

AI Summary

Data-driven trajectory prediction models for robotics suffer from degraded reliability, miscalibrated uncertainty estimates, and physical inconsistency under out-of-distribution (OOD) observations. Method: We propose a robust learning–physics hybrid prediction framework featuring a plug-and-play switching mechanism: conformal prediction detects dynamical anomalies in real time and triggers an automatic switch to a physically constrained distributed port-Hamiltonian system (dPHS); crucially, we embed Gaussian processes into the dPHS energy function, enabling joint optimization of dynamics learning and Bayesian uncertainty quantification. Contribution/Results: Experiments demonstrate that our approach significantly improves prediction reliability under OOD conditions, yields well-calibrated uncertainty intervals, and ensures physical consistency and decision safety—thereby bridging the gap between data-driven flexibility and physics-based trustworthiness in robotic trajectory forecasting.

Technology Category

• Intelligent Robots: State Estimation
• Humans and AI: Human-Aware Planning and Behavior Prediction
• Machine Learning: Calibration & Uncertainty Quantification

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

The ability to predict trajectories of surrounding agents and obstacles is a crucial component in many robotic applications. Data-driven approaches are commonly adopted for state prediction in scenarios where the underlying dynamics are unknown. However, the performance, reliability, and uncertainty of data-driven predictors become compromised when encountering out-of-distribution observations relative to the training data. In this paper, we introduce a Plug-and-Play Physics-Informed Machine Learning (PnP-PIML) framework to address this challenge. Our method employs conformal prediction to identify outlier dynamics and, in that case, switches from a nominal predictor to a physics-consistent model, namely distributed Port-Hamiltonian systems (dPHS). We leverage Gaussian processes to model the energy function of the dPHS, enabling not only the learning of system dynamics but also the quantification of predictive uncertainty through its Bayesian nature. In this way, the proposed framework produces reliable physics-informed predictions even for the out-of-distribution scenarios.

Problem

• Predicting trajectories of surrounding agents reliably
• Addressing performance issues in out-of-distribution scenarios
• Combining data-driven and physics-based models for uncertainty

Innovation

• Plug-and-Play Physics-Informed Machine Learning framework
• Conformal prediction for outlier dynamics identification
• Gaussian processes for uncertainty quantified Port-Hamiltonian models