Using Clarke Transform to Create a Framework on the Manifold: From Sampling via Trajectory Generation to Control

Reinhard M. Grassmann, Jessica Burgner-Kahrs · 2024

Read Using Clarke Transform to Create a Framework on the Manifold: From Sampling via Trajectory Generation to Control by Reinhard M. Grassmann, Jessica Burgner-Kahrs (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2412-16422.txt.

Introduction

Reinhard Grassmann and Jessica Burgner-Kahrs give a frame for soft robots that bend, built on a change of coordinates named for Clarke. The frame has three parts that work on a curved space of shapes: a planner, a path maker, and a controller.

All parts are fast, small, and free of branches in the code. A coder lets older parts that use joint angles plug in. The authors tie the joint limits to limits on the curved space, so paths stay smooth there. They also link a bend limit to curves that run side by side. A sim of a four part soft robot shows the frame at work.

Catalog record

CREATORReinhard M. Grassmann, Jessica Burgner-Kahrs
YEAR2024
COLLECTIONPAPERS
FORMATTXT
FILE SIZE41 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2412-16422

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Using Clarke Transform to Create a Framework on the Manifold: From Sampling via Trajectory Generation to Control Reinhard M. Grassmann, Jessica Burgner-Kahrs Abstract We present a framework based on Clarke coordinates for spatial displacement-actuated continuum robots with an arbitrary number of joints. This framework consists of three modular components, i.e. , a planner, trajectory generator, and controller defined on the manifold. All components are computationally efficient, compact, and branchless, and an encoder can be used to interface existing framework components that are not based on Clarke coordinates. We derive the relationship between the kinematic constraints in the joint space and on the manifold to generate smooth trajectories on the manifold. Furthermore, we establish the connection between the displacement constraint and parallel curves. To demonstrate its effectiveness, a demonstration in simulation for a displacement-actuated continuum robot with four segments is presented. I Introduction Soft and continuum robots must necessarily move from a current configuration to a desired configuration to accomplish meaningful real-world tasks in medical and industrial applications. Thus, a minimal framework should consist of three parts; planner, trajectory generator, and controller. However, developed frameworks are tailored to a specific robot morphology rendering some, if not all, components of the framework unsuitable for other robot morphologies. This limits the reusability and rapid advancement in the field. Overcoming that by providing a modular framework, where each component has the same interface, has obvious benefits, e.g. , (i) interpretability and safety due to a divide-and-conquer approach, (ii) compatibility due to the same interface, (iii) flexibility in changing and testing components, (vi) agility in developing, improving, and testing of separate components, and (v) empowering the field by sharing and reusing components. Overall, an increase in cost-efficiency and a decrease in the number of times the wheel needs to be reinvented can be expected. Therefore, providing a general principle to create frameworks is a necessary step to move forward together, rather than just improving one particular robot morphology. As a candidate for an interface between the components, different improved state representations [ 1 , 2 , 3 , 4 , 5 ] have been introduced in recent years. It has been shown that using improved state representations has decisive advantages. Della Santina et al. [ 1 ] discuss and analyze the main issues arising from the use of commonly used parameterization, i.e. , curvature \kappa and bending plane angle \theta used for constant curvature models. The derived improved state representation [ 1 ] for n=4 number of joints shows superb performance on a model-based control application. Allen et al. [ 2 ] introduces improved state representations for n=3 and n=4 number of joints. They propose closed-form and singularity-free forward kinematics as well as Jacobian-based inverse kinematics mitigating the disadvantages of previous approaches. Dian et al. [ 4 ] proposed an improved state representation that can be seen as an adaptation of improved state representation by Della Santina et al. [ 1 ] to n=3 . As mentioned by Dupont et al. [ 6 ] , different representations exist and among them, that utilize Cartesian curvature components, i.e. , \kappa_{x} and \kappa_{y} , have the advantage of avoiding a parametric singularity. The connection between the Cartesian curvature components and all improved state representations is established by Grassmann et al. [ 5 ] . Grassmann et al. [ 5 ] links the displacement-actuated joints and the displacement constraint to electrical currents and Kirchhoff’s current law, respectively. This and the reformulation of the joint representation lead to the use of the Clarke transformation matrix that allows the disentanglement of the displacement constraint. As a result, an improved state representation called Clarke coordinates is derived unifying previous improved state representations [ 1 , 2 , 3 , 4 , 5 ] and generalizing them to n\geq 3 . Since constant curvature [ 7 ] is not enforced during their derivation, the Clarke coordinates are not limited to constant curvature assumption. Grassmann & Burgner-Kahrs [ 8 ] show that the analogy to the Clarke transformation matrix is not necessary to derive the so-called Clarke transform. This allows for consideration of arbitrary joint location distribution on the constant cross-section of a displacement-actuated continuum robot. In general, the derived approaches based on improved state representations are closed-from, compact, computationally efficient, and singularity-free. This makes the Clarke coordinates a desirable representation and unified language to interface all components within a framework. Due to the recent introduction of the Clarke transform, only sampling methods [ 5 , 8 ] and controllers [ 5 , 8 ] have been proposed. While previous sampling methods and controllers can be used as components in a framework, a trajectory generator has yet to be formulated using Clarke coordinates. This work considers a displacement-actuated continuum robot defined by kinematic design parameters consisting of segment length l=\textit{const.} , distance d_{i} , angle \psi_{i} , and number of joints n . We kindly refer to [ 5 , 8 ] for a depiction. For n equally distributed joints, the polar coordinate given by \displaystyle\psi_{i}=\dfrac{2\pi}{n}(i-1)\quad\text{and}\quad d_{i}=d=\textit{const.} (1) describes the location of the i\textsuperscript{th} joint in the cross-section. The kinematic design parameters fully describe the displacement-actuated continuum robot in the kinematic sense encompassing a large class of soft robots and continuum robots. In the presented work, we aim to shift the paradigm towards a unified approach by introducing a framework, where each of the components is based on the Clarke coordinates of a 2\text{\,}\mathrm{d}\mathrm{o}\mathrm{f} manifold. In particular, the following contributions are made: • Presentation of a framework completely based on the Clarke coordinates • Introduction of a smooth trajectory generator that specifies and generates trajectories on the 2\text{\,}\mathrm{d}\mathrm{o}\mathrm{f} manifold • Derivation of the kinematics constraints for trajectories on the manifold • Connection between displacement constraint and parallel curves also known as offset curves II Clarke Transform In this section, we provide a self-contained and brief description of the Clarke transform, the relationship to arc parameters, and virtual displacement. The description of virtual displacement is expanded to the connection between the displacement constraint and parallel curves. Furthermore, we briefly state the encoder-decoder architecture. For a longer discussion, we kindly refer to [ 5 , 8 , 9 ] . II-A Clarke Transform for Displacement-Actuated Joint Proposed by Grassmann at al. [ 5 ] , a vector with values of each displacement-actuated joint \rho_{i} can be represented as \displaystyle\boldsymbol{\rho}=\begin{bmatrix}\rho_{1}\\ \rho_{2}\\ \vdots\\ \rho_{n}\end{bmatrix}=\begin{bmatrix}\rho_{\mathrm{Re}}\cos\left(\psi_{1}\right)+\rho_{\mathrm{Im}}\sin\left(\psi_{1}\right)\\ \rho_{\mathrm{Re}}\cos\left(\psi_{2}\right)+\rho_{\mathrm{Im}}\sin\left(\psi_{2}\right)\\ \vdots\\ \rho_{\mathrm{Re}}\cos\left(\psi_{n}\right)+\rho_{\mathrm{Im}}\sin\left(\psi_{n}\right)\end{bmatrix}\in\mathbf{Q}, (2) where \mathbf{Q}\subset\mathbb{R}^{n} is the joint space. All joint values in ( 2 ) are interdependent and constrained obeying the displacement constraint given by \displaystyle\sum_{i=1}^{n}\rho_{i}=0. (3) The Clarke coordinates [ 5 ] , i.e. , \rho_{\mathrm{Re}} and \rho_{\mathrm{Im}} in ( 2 ), can be combined into a vector as well, i.e. , \displaystyle\overline{\boldsymbol{\rho}}=\begin{bmatrix}\rho_{\mathrm{Re}}\\ \rho_{\mathrm{Im}}\end{bmatrix}. (4) Both representations can be transformed into each other via \displaystyle\overline{\boldsymbol{\rho}} \displaystyle=\boldsymbol{M}_{\mathcal{P}}\boldsymbol{\rho}\quad\text{and}\quad (5) \displaystyle\boldsymbol{\rho} \displaystyle=\boldsymbol{M}_{\mathcal{P}}^{-1}\overline{\boldsymbol{\rho}}, (6) where the generalized Clarke transformation matrix \boldsymbol{M}_{\mathcal{P}} is \displaystyle\boldsymbol{M}_{\mathcal{P}}=\dfrac{2}{n}\!\begin{bmatrix}\cos\left(\psi_{0}\right)&\cos\left(\psi_{1}\right)&\cdots&\cos\left(\psi_{n}\right)\\ \sin\left(\psi_{0}\right)&\sin\left(\psi_{1}\right)&\cdots&\sin\left(\psi_{n}\right)\end{bmatrix}, (7) which holds for symmetrically arranged joints [ 5 , 8 ] . By factor out ( 4 ) in ( 2 ), its inverse \boldsymbol{M}_{\mathcal{P}}^{-1} can be found being \displaystyle\boldsymbol{M}_{\mathcal{P}}^{-1}=\begin{bmatrix}\cos\left(\psi_{1}\right)&\sin\left(\psi_{1}\right)\\ \cos\left(\psi_{2}\right)&\sin\left(\psi_{2}\right)\\ \vdots&\vdots\\ \cos\left(\psi_{n}\right)&\sin\left(\psi_{n}\right)\end{bmatrix}, (8) where ( 8 ) is the right-inverse of non-square matrix ( 7 ). II-B Arc Parameters Utilizing the constant curvature assumption [ 7 ] ,

Read the complete text