Linearization and Identification of Multiple-Attractors Dynamical System through Laplacian Eigenmaps

Abstract

Dynamical Systems (DS) are fundamental to the modeling and understanding of time evolving phenomena, and find application in physics, biology and control. As determining an analytical description of the dynamics is often difficult, data-driven approaches are preferred for identifying and controlling nonlinear DS with multiple equilibrium points. Identification of such DS has been treated largely as a supervised learning problem. Instead, we focus on a unsupervised learning scenario where we know neither the number nor the type of dynamics. We propose a Graph-based spectral clustering method that takes advantage of a velocity-augmented kernel to connect data-points belonging to the same dynamics, while preserving the natural temporal evolution. We study the eigenvectors and eigenvalues of the Graph Laplacian and show that they form a set of orthogonal embedding spaces, one for each sub-dynamics. We prove that there always exist a set of 2-dimensional embedding spaces in which the sub-dynamics are linear, and n-dimensional embedding where they are quasi-linear. We compare the clustering performance of our algorithm to Kernel K-Means, Spectral Clustering and Gaussian Mixtures and show that, even when these algorithms are provided with the true number of sub-dynamics, they fail to cluster them correctly. We learn a diffeomorphism from the Laplacian embedding space to the original space and show that the Laplacian embedding leads to good reconstruction accuracy and a faster training time through an exponential decaying loss, compared to the state of the art diffeomorphism-based approaches.

Learning High Dimensional Demonstrations Using Laplacian Eigenmaps

“…Therefore, we obtain a linear embedding of the dynamics in the latent space. This result appears in [17].…”

“…In most common robotic applications, such as pick and place objects, the task space is higher than two dimensions. This paper extends the proposed algorithm in [17] to learn complex demonstrations in high dimensions given a single demonstration for training.…”

“…In [17], the problem has been addressed for a demonstration evolving in a 2D space. Therefore, the method is only applicable to learn a demonstration in joint space for a robot with two joints or to learn a demonstration in a 2-dimensional task space.…”

“…The n smallest repeating eigenvalues belong to the order O(1/N 2 ) from (17). The eigenvectors of L(G) under consideration are same as those of B i and satisfy its eigenequation, therefore, We suppress the dependency on i for simplicity and denote u i = [u i 1 , .…”