Parameter-Robust MPPI for Safe Online Learning of Unknown Parameters

2026-01-06

IEEE Robotics and Automation Letters

AI Summary

This work addresses the challenge of balancing safety and performance in robotic control under dynamic environments with uncertain or time-varying physical parameters. To this end, the authors propose a Parameter-Robust Model Predictive Path Integral (PRMPPI) control framework that uniquely integrates Stein variational gradient descent with conformal prediction to enable online belief updates over model parameters and probabilistic safety verification. The approach simultaneously optimizes a performance-oriented primary trajectory and a safety-oriented backup trajectory, effectively balancing exploration and safety. Leveraging a particle-filter-like representation of parameter uncertainty, the method demonstrates significant improvements over baseline approaches in both simulation and hardware experiments, achieving higher task success rates, lower tracking errors, and more accurate parameter estimation.

Technology Category

• Intelligent Robots: State Estimation
• Natural Language Processing: Safety and Robustness
• Reasoning under Uncertainty: Stochastic Optimization

Application Category

• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Security and Privacy: Security and privacy of machine learning and AI applications

Abstract

Robots deployed in dynamic environments must remain safe even when key physical parameters are uncertain or change over time. We propose Parameter-Robust Model Predictive Path Integral (PRMPPI) control, a framework that integrates online parameter learning with probabilistic safety constraints. PRMPPI maintains a particle-based belief over parameters via Stein Variational Gradient Descent, evaluates safety constraints using Conformal Prediction, and optimizes both a nominal performance-driven and a safety-focused backup trajectory in parallel. This yields a controller that is cautious at first, improves performance as parameters are learned, and ensures safety throughout. Simulation and hardware experiments demonstrate higher success rates, lower tracking error, and more accurate parameter estimates than baselines.

Problem

• parameter uncertainty
• online learning
• safety constraints
• robotic control
• dynamic environments

Innovation

• Parameter-Robust MPPI
• Online Parameter Learning
• Conformal Prediction
• Stein Variational Gradient Descent
• Safety-Critical Control