Enhancing Trajectory Tracking in Humanoid Robots Using Neural Network-Based Dynamic Gain Control

Abstract

This paper presents the development and evaluation of a dynamic gain controller utilizing neural networks to enhance trajectory tracking performance in the NAO humanoid robot. The proposed controller employs a differential kinematic model and dynamically adjusts its gains using a backpropagation algorithm, eliminating the need for manual gain tuning and simplifying the robot's setup process. Experimental validation was conducted in a simulated environment using CoppeliaSim, with the NAOqi library facilitating integration. The analysis results demonstrate that the dynamic controller using a neural network provides better trajectory tracking accuracy than the traditional kinematic controller. Adaptability of the dynamic controller, which adjusts gain parameters in real-time, contributes to improved robustness and precision across various trajectory types. These findings demonstrate the potential of dynamic, self-tuning controllers in enhancing the performance, efficiency, and versatility of humanoid robots in complex navigation tasks. Doi: 10.28991/ESJ-2025-09-02-02 Full Text: PDF