Stable Adaptive Neural Filters for Teleoperations with Uncertain Delays
ICRA 2023 presentation and IEEE RA‑L publication on adaptive neural filters that stabilize teleoperation systems under unknown and time‑varying communication delays.
Project Highlights
This project introduces an adaptive neural filter to improve stability and performance in teleoperation systems under random, time‑varying delays. Key challenges addressed:
- Instability from network uncertainties: Random delays in communication networks can destabilize teleoperation systems.
- Degraded performance: Delays reduce operator transparency and system usability.
- Balancing stability and performance: Conventional methods are often overly conservative, limiting performance.
Neural Filter Approach
An adaptive neural network dynamically adjusts filter coefficients, ensures stability via Lyapunov‑Krasovskii analysis, avoids local minima, and provides accurate signal estimations under uncertain delays.
Experimental results: Demonstrated outstanding performance, improved stability, and smoother teleoperation operation on a real‑world platform.