Uncertainty-Aware Predictive Safety Filters for Probabilistic Neural Network Dynamics

2026-04-29

AI Summary

Existing predictive safety filters rely on first-principles models or Gaussian processes, which struggle to scale to high-dimensional systems. Meanwhile, dynamics models based on probabilistic ensemble neural networks lack rigorous uncertainty quantification, compromising safety guarantees. This work proposes the Uncertainty-aware Predictive Safety Filter (UPSi), which for the first time integrates rigorous uncertainty quantification into a probabilistic ensemble framework. UPSi characterizes safe state evolution via reachable sets and incorporates explicit confidence constraints to prevent unsafe model extrapolation. Evaluated on standard safe reinforcement learning benchmarks, the method significantly enhances exploration safety while maintaining task performance comparable to state-of-the-art model-based reinforcement learning approaches, thereby enabling scalable model predictive control with formal safety assurances.

Abstract

Predictive safety filters (PSFs) leverage model predictive control to enforce constraint satisfaction during deep reinforcement learning (RL) exploration, yet their reliance on first-principles models or Gaussian processes limits scalability and broader applicability. Meanwhile, model-based RL (MBRL) methods routinely employ probabilistic ensemble (PE) neural networks to capture complex, high-dimensional dynamics from data with minimal prior knowledge. However, existing attempts to integrate PEs into PSFs lack rigorous uncertainty quantification. We introduce the Uncertainty-Aware Predictive Safety Filter (UPSi), a PSF that provides rigorous safety predictions using PE dynamics models by formulating future outcomes as reachable sets. UPSi introduces an explicit certainty constraint that prevents model exploitation and integrates seamlessly into common MBRL frameworks. We evaluate UPSi within Dyna-style MBRL on standard safe RL benchmarks and report substantial improvements in exploration safety over prior neural network PSFs while maintaining performance on par with standard MBRL. UPSi bridges the gap between the scalability and generality of modern MBRL and the safety guarantees of predictive safety filters.

Problem

• Predictive Safety Filters
• Probabilistic Neural Networks
• Uncertainty Quantification
• Model-based Reinforcement Learning
• Safe Exploration

Innovation

• Predictive Safety Filter
• Probabilistic Ensemble
• Uncertainty Quantification
• Model-Based Reinforcement Learning
• Reachable Sets