Dynamical Systems (DS) are crucial for robotics control, ensuring stability and adaptability. Current strategies face trade-offs between stability and complexity. A new method encodes non-linearity within space curvature, enhancing DS complexity efficiently. The approach integrates local non-linearities seamlessly, showcasing obstacle avoidance through harmonic damped oscillators on a latent manifold. By learning the manifold's embedded representation, non-linearity is encoded within space curvature, enabling direct induction of local deformations for obstacle avoidance. The methodology demonstrates effectiveness in 2D synthetic vector fields and 3D robotic end-effector motions.
To Another Language
from source content
arxiv.org
Principais Insights Extraídos De
by Bernardo Fic... às arxiv.org 03-19-2024
https://arxiv.org/pdf/2403.11948.pdfPerguntas Mais Profundas