Parametric Modeling of Deformable Linear Objects for Robotic Outfitting and Maintenance of Space Systems

Abstract

Outfitting and maintenance are important to an in-space architecture consisting of long duration missions.During such missions, crew is not continuously present; robotic agents become essential to the construction, maintenance, and servicing of complicated space assets, requiring some degree of autonomy to plan and execute tasks.There has been significant research into manipulation planning for rigid elements for in-space assembly and servicing, but outfitting and maintenance of flexible electrical cables, which fall under the domain of Deformable Linear Objects (DLOs), have not received such attention despite being critical components of powered space systems. Cables often have a non-zero bend equilibrium configuration, which the majority of DLO research does not consider. This dissertation implements a model-based optimization approach to estimate cable configuration, where a design parameter of the model's discretization level enables trading model accuracy vs. computational complexity. Such a consideration when creating the DLO model has not been well explored in the domain of DLO modeling. Observed 2D cable configurations are used to improve the model via parameter estimation of the equilibrium configurations, under multiple loading and constraint conditions. The parameter estimation is formulated by augmenting the optimization problem used to solve for the configuration to weight the solution towards one that also minimizes error to the measured DLO center-line. The parameters can then be adjusted through a least squares solution across all collected data with respect to the parameters required for static equilibrium. The parameter estimation is validated through comparing predicted configurations based on estimated parameters to that of a real cable. The incorporation of parameter estimation to the cable model is shown to reduce prediction errors. A dynamic formulation is implemented to allow the parametrically built DLO to be manipulated in a constrained environment. The model is validated against a real dynamic cable manipulation with a UR10e robot arm, both with and without the estimated parameters determined through static configurations. The results of this work demonstrate some of the challenges present with robotic cable manipulation, exploring the complexities of outfitting and maintenance operations of in-space facilities, and puts forth a method for reducing the size of the state space of a cable payload while accounting for non-zero equilibrium configurations.