Adaptive Model Learning of Neural Networks with UUB Stability for Robot Dynamic Estimation

TL;DR

This paper introduces an adaptive neural network-based method for real-time robot dynamic estimation, ensuring stability and improved accuracy through Lyapunov-based learning rules, demonstrated on a 3-DOF haptic device.

Contribution

It presents a novel stable adaptive learning scheme for neural networks using Lyapunov functions, enabling online robot dynamic identification with guaranteed stability.

Findings

Effective dynamic estimation on a 3-DOF haptic device

Stable neural network updating rules derived from Lyapunov functions

Improved robustness over batch algorithms in dynamic estimation

Abstract

Since batch algorithms suffer from lack of proficiency in confronting model mismatches and disturbances, this contribution proposes an adaptive scheme based on continuous Lyapunov function for online robot dynamic identification. This paper suggests stable updating rules to drive neural networks inspiring from model reference adaptive paradigm. Network structure consists of three parallel self-driving neural networks which aim to estimate robot dynamic terms individually. Lyapunov candidate is selected to construct energy surface for a convex optimization framework. Learning rules are driven directly from Lyapunov functions to make the derivative negative. Finally, experimental results on 3-DOF Phantom Omni Haptic device demonstrate efficiency of the proposed method.