Abstract:In numerous robotics and mechanical engineering applications, among others, data is often constrained on smooth manifolds due to the presence of rotational degrees of freedom. Common datadriven and learning-based methods such as neural ordinary differential equations (ODEs), however, typically fail to satisfy these manifold constraints and perform poorly for these applications. To address this shortcoming, in this paper we study a class of neural ordinary differential equations that, by design, leave a given manifold invariant, and characterize their properties by leveraging the controllability properties of control affine systems. In particular, using a result due to Agrachev and Caponigro on approximating diffeomorphisms with flows of feedback control systems, we show that any map that can be represented as the flow of a manifold-constrained dynamical system can also be approximated using the flow of manifold-constrained neural ODE, whenever a certain controllability condition is satisfied. Additionally, we show that this universal approximation property holds when the neural ODE has limited width in each layer, thus leveraging the depth of network instead for approximation. We verify our theoretical findings using numerical experiments on PyTorch for the manifolds S2 and the 3-dimensional orthogonal group SO(3), which are model manifolds for mechanical systems such as spacecrafts and satellites. We also compare the performance of the manifold invariant neural ODE with classical neural ODEs that ignore the manifold invariant properties and show the superiority of our approach in terms of accuracy and sample complexity.

Controller Design for Systems on Manifolds in Euclidean Space

On Controller Design for Systems on Manifolds in Euclidean Space

Control Via Interconnection and Damping Assignment of Linear Time-Invariant Systems: a Tutorial

A Controller-Design Method For Nonlinear-Systems Via A Staged-Transformation Technique

Model Predictive Tracking Control for Invariant Systems on Matrix Lie Groups via Stable Embedding Into Euclidean Spaces

Model Predictive Regulation on Manifolds in Euclidean Space

Immersion and Invariance Stabilization of Nonlinear Systems: A Horizontal Contraction Approach

A Generalized Trajectory Tracking Controller for Robot Manipulators with Bounded Inputs

Almost Global Asymptotic Trajectory Tracking for Fully-Actuated Mechanical Systems on Homogeneous Riemannian Manifolds

Model Predictive Control for Trajectory Tracking on Differentiable Manifolds

Control problems in super-articulated mechanical systems

An improved Udwadia–Kalaba approach for controller design in underactuated mechanical systems

Unifying Controller Design for Stabilizing Nonlinear Systems with Norm-Bounded Control Inputs

Integral Control On Nonlinear Spaces: Two Extensions

Tracking control for a class of fully actuated systems on manifolds with impulsive effects

MPC on manifolds with an application to the control of spacecraft attitude on SO(3)

Learning on Manifolds: Universal Approximations Properties using Geometric Controllability Conditions for Neural ODEs

Control contraction metrics on Lie groups

Stabilizing the Second-Order Nonholonomic Systems with Chained Form by Finite-Time Stabilizing Controllers

Point Stabilization and Trajectory Tracking of Underactuated Surface Vessels: A Geometric Control Approach.

New Stabilization Controller of State-Constrained Nonholonomic Systems With Disturbances: Theory and Experiment