Learning the Kinematics of Tubular Continuum Robots: Model-based vs. Data-based Methods

学习管状连续体机器人的运动学：基于模型与基于数据的方法

• 批准号: RGPIN-2019-04846
• 负责人: BurgnerKahrs, Jessica
• 金额: $ 3.86万
• 依托单位: University of Toronto
• 依托单位国家: 加拿大
• 项目类别: Discovery Grants Program - Individual
• 财政年份: 2022
• 资助国家: 加拿大
• 起止时间: 2022-01-01 至 2023-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=758282
• 关键词: Learning; Kinematics; Tubular; Continuum; Robots

Tubular continuum robots are the smallest among all continuum robots with typical diameter to length ratios of 1:250. The composition of concentrically arranged elastic tubes with pre--curvatures allows for simple, yet dextrous, robotic manipulators at a millimetre scale. Actuation is achieved mechanically by relative rotation and translation of tubes and results in tentacle--like motions. Tubular continuum robots are particularly promising for medical interventions, such as diagnosis through natural orifices or minimally invasive surgery. For instance, they could reach the skull base through the nasal passage to remove tumours or deliver drugs on curvilinear paths to locations deep within the human body. Despite the simple actuation of component tubes, the resulting motion of tubular continuum robots is characterised by a highly non-linear behaviour due to elastic interactions. Today, the modelling, computational design, and motion planning methodologies for those robots are characterised by limitations in terms of accuracy and efficiency. The gold standard approaches are a trade-off between the consideration of a limited number of physical parameters and computational expense. As a result, tubular continuum robots have not left the laboratory bench-top despite all their potential merits in health care. This research programme aims in leveraging data-based approaches and deep learning techniques for modelling, computational design, and motion planning. The proposed research is structured around four scientific key questions: 1) How can the curvilinear structure, the morphological constraints, and the mechanical laws that govern tubular continuum robots in the real world be leveraged and exploited by deep learning? 2) How can data be generated, experimentally obtained, and represented for tubular continuum robots in order to inform learning-based approaches? 3) How can learning-based approaches help to exploit the relevant parameter space for physics-based models of tubular continuum robots? 4) How great is the potential to overcome the limitations of current model-based approaches with data-driven problem solving? To the best of my knowledge, learning-based approaches are proposed for the first time for tubular continuum robots by my research group. Deep learning can serve to discover unknown problem structures and to derive novel knowledge, which can then be used to improve and expand existing problem-specific algorithms. The proposed programme will generate scientific methodologies for tubular continuum robots in particular as well as soft and continuum robotics in general. It will further contribute to the understanding of relevant physical phenomena for model-based solutions. The expected step change research results will ultimately serve as an enabler to exploit the full potential of tubular continuum robots in medical interventions.

管状连续体机器人是所有连续体机器人中最小的，典型的直径与长度比为1：250。具有预曲率的同心排列的弹性管的组成允许在毫米尺度上使用简单而灵活的机器人操纵器。驱动是通过管子的相对旋转和平移来机械实现的，并导致触手状的运动。管状连续体机器人在医疗干预方面特别有希望，例如通过自然开口进行诊断或微创手术。例如，它们可以通过鼻腔到达颅底，以移除肿瘤，或者通过曲线路径将药物输送到人体深处。尽管组件管的驱动很简单，但由于弹性相互作用，管状连续体机器人的运动具有高度非线性的特点。今天，这些机器人的建模、计算设计和运动规划方法的特点是在精度和效率方面存在局限性。黄金标准方法是考虑有限数量的物理参数和计算成本之间的权衡。因此，尽管管状连续体机器人在医疗保健方面具有所有潜在的优点，但它们并没有离开实验室的工作台。这项研究计划旨在利用基于数据的方法和深度学习技术来进行建模、计算设计和运动规划。所提出的研究围绕四个科学关键问题展开：1)管状连续体机器人在现实世界中的曲线结构、形态约束和力学规律如何被深度学习所利用？2)如何为管状连续体机器人生成、实验获取和表示数据以提供基于学习的方法？3)基于学习的方法如何帮助开发管状连续体机器人基于物理模型的相关参数空间？4)通过数据驱动解决问题来克服当前基于模型的方法的限制的潜力有多大？据我所知，我的研究小组首次针对管状连续体机器人提出了基于学习的方法。深度学习可以用来发现未知的问题结构和获得新的知识，然后可以用来改进和扩展现有的特定于问题的算法。拟议方案将特别为管状连续体机器人以及一般的软机器人和连续体机器人产生科学方法。它将进一步有助于理解基于模型的解决方案的相关物理现象。预期的阶跃变化研究结果最终将成为开发管状连续体机器人在医疗干预中的全部潜力的推动者。