An optimization-based motion planner for dual-arm manipulation of the soft deformable linear objects with nonnegligible gravity

Abstract

The dual-arm manipulation of deformable linear objects (DLOs) represents a practical and challenging problem in robotics research, offering significant potential for various industrial applications, including cable assembly. To accurately model the mechanical properties of DLOs, a Kirchhoff differential model is employed, which parameterizes the DLO configuration as a 6-dimensional manifold. Traditionally, approaches to solving this planning problem relied solely on sampling-based methods, incurring high computational costs due to the necessity of obtaining the DLO shape for each sample. Additionally, these methods completely ignored gravity, assuming that the DLO was stiff enough. However, in many industrial scenarios, this assumption cannot hold, particularly when dealing with soft DLOs, where the effects of gravity are non-negligible, leading to poorer stability and sensitivity. In this work, a novel optimization-based paradigm is proposed for the manipulation planning of soft DLOs with dual arms, addressing the challenges associated with their soft nature and the influence of gravity. The concept of ’stability distance’ is introduced as an easily measurable indicator of the degree of DLO stability. Furthermore, a thorough investigation into the singularity phenomenon in DLO local leading is conducted to identify its causes and provide effective solutions. Additionally, a strategy is introduced to avoid local traps of the DLO in complex obstacle environments. The comprehensive planner is validated through both simulation and hardware experiments, utilizing two types of soft DLOs with a length of approximately 1 m in various environmental settings. The results demonstrate the promising performance of the algorithm across diverse assembly scenarios.