Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Gated Uncertainty-Aware Runtime Dual Invariants for Neural Signal-Controlled Robotics

Tasha Kim, Oiwi Parker Jones

TL;DR

This work addresses safety in neural signal-based robotics by introducing GUARDIAN, a calibration-aware, neuro-symbolic runtime verification framework that links EEG-derived intent to symbolic goals via a groundable plan. It combines a calibrated intent distribution, entropy/artifact/oscillation sensing, and a dual-layer monitor that enforces both physiological and logical invariants, yielding auditable safety traces. The approach achieves 94–97% safety even when decoders perform poorly (27–46% test accuracy) and maintains sub-millisecond decision latency at 100 Hz, with a 1.7× increase in correct interventions under simulated noise. This provides a practical, decoder-agnostic safety guard suitable for real-time assistive robotics and aligns with regulatory expectations through its auditable neuro-symbolic pipeline and minimal overhead.

Abstract

Safety-critical assistive systems that directly decode user intent from neural signals require rigorous guarantees of reliability and trust. We present GUARDIAN (Gated Uncertainty-Aware Runtime Dual Invariants), a framework for real-time neuro-symbolic verification for neural signal-controlled robotics. GUARDIAN enforces both logical safety and physiological trust by coupling confidence-calibrated brain signal decoding with symbolic goal grounding and dual-layer runtime monitoring. On the BNCI2014 motor imagery electroencephalogram (EEG) dataset with 9 subjects and 5,184 trials, the system performs at a high safety rate of 94-97% even with lightweight decoder architectures with low test accuracies (27-46%) and high ECE confidence miscalibration (0.22-0.41). We demonstrate 1.7x correct interventions in simulated noise testing versus at baseline. The monitor operates at 100Hz and sub-millisecond decision latency, making it practically viable for closed-loop neural signal-based systems. Across 21 ablation results, GUARDIAN exhibits a graduated response to signal degradation, and produces auditable traces from intent, plan to action, helping to link neural evidence to verifiable robot action.

Gated Uncertainty-Aware Runtime Dual Invariants for Neural Signal-Controlled Robotics

TL;DR

This work addresses safety in neural signal-based robotics by introducing GUARDIAN, a calibration-aware, neuro-symbolic runtime verification framework that links EEG-derived intent to symbolic goals via a groundable plan. It combines a calibrated intent distribution, entropy/artifact/oscillation sensing, and a dual-layer monitor that enforces both physiological and logical invariants, yielding auditable safety traces. The approach achieves 94–97% safety even when decoders perform poorly (27–46% test accuracy) and maintains sub-millisecond decision latency at 100 Hz, with a 1.7× increase in correct interventions under simulated noise. This provides a practical, decoder-agnostic safety guard suitable for real-time assistive robotics and aligns with regulatory expectations through its auditable neuro-symbolic pipeline and minimal overhead.

Abstract

Safety-critical assistive systems that directly decode user intent from neural signals require rigorous guarantees of reliability and trust. We present GUARDIAN (Gated Uncertainty-Aware Runtime Dual Invariants), a framework for real-time neuro-symbolic verification for neural signal-controlled robotics. GUARDIAN enforces both logical safety and physiological trust by coupling confidence-calibrated brain signal decoding with symbolic goal grounding and dual-layer runtime monitoring. On the BNCI2014 motor imagery electroencephalogram (EEG) dataset with 9 subjects and 5,184 trials, the system performs at a high safety rate of 94-97% even with lightweight decoder architectures with low test accuracies (27-46%) and high ECE confidence miscalibration (0.22-0.41). We demonstrate 1.7x correct interventions in simulated noise testing versus at baseline. The monitor operates at 100Hz and sub-millisecond decision latency, making it practically viable for closed-loop neural signal-based systems. Across 21 ablation results, GUARDIAN exhibits a graduated response to signal degradation, and produces auditable traces from intent, plan to action, helping to link neural evidence to verifiable robot action.

Paper Structure

This paper contains 51 sections, 13 equations, 4 figures, 15 tables, 4 algorithms.

Table of Contents

  1. Introduction
  2. GUARDIAN: Formalization and Algorithm
  3. Calibration-aware safety.
  4. Experimental Setup
  5. Results and Analysis
  6. Discussion
  7. Conclusion
  8. Safety Definitions and Metrics
  9. Dataset Specifications
  10. BNCI2014_001 dataset details.
  11. Preprocessing Pipeline
  12. Algorithm Details
  13. Decoder Architecture Details
  14. EEGNet
  15. Riemannian Decoder
  16. ...and 36 more sections

Figures (4)

  • Figure 1: System architecture overview.
  • Figure 2: Confusion matrix for EEGNet (test set).
  • Figure 3: Optimal thresholds under different objective weights. Plot shows safety-intervention trade-off (including F1-Optimal), indicating the Pareto frontier that maximizes safety and minimizes interventions.
  • Figure 4: Component latency comparison measured across decoders (mean $\pm$ SD, n=10,000 trials).