Robotic skill transfer and augmentation for contact-rich tasks in manufacturing (STAMAN)

制造中接触丰富的任务的机器人技能转移和增强 (STAMAN)

• 批准号: EP/Y02270X/1
• 负责人: Yongjing Wang
• 金额: $ 125.8万
• 依托单位: University of Birmingham
• 依托单位国家: 英国
• 项目类别: Research Grant
• 财政年份: 2023
• 资助国家: 英国
• 起止时间: 2023 至 无数据
• 项目状态: 未结题
• 来源: https://gtr.ukri.org/projects?ref=EP%2FY02270X%2F1
• 关键词: Robotic; skill; transfer; augmentation; contact

Many assembly and disassembly tasks in manufacturing have small clearances and limited accessibility, such as shaft-hole insertion/separation and bolt-nut assembly/disassembly. Using robots in these contact-rich tasks is more complex than those having no physical contacts (e.g. computer visual inspection) or simple contacts (e.g. cutting, welding, pick-and-place). The deployment of robots in contact-rich tasks has been limited to date. The contact-rich tasks that involve complex shapes, small clearances or deformable materials are particularly challenging to robotise due to the likely events of jamming and wedging. Our previous research has investigated techniques that allow robots to learn contact-rich skills (e.g. complex motion plans and force control policies) using two main AI-based pathways: (1) self-learning from trial-and-error, and (2) learning from human demonstrations. The two participating universities, Birmingham and Sheffield, have research experiences in (1) and (2), respectively.A key challenge observed in the current research is that in many cases a robot's contact-rich skill cannot be performed by other robots of different motion properties (e.g. accuracy, precision and stiffness), or be applied to a new task with variations (e.g. differences in object geometry, shape, and materials). This is because a robotic contact-rich skill, i.e. control policies and motion plans, is usually acquired for a specific task and cannot be adopted by new robots or in new tasks.STAMAN's vision is to create AI-based mechanisms to allow robots to share and recreate obtained digital skills (e.g. motion and force/torque control strategies) to allow easy automation scale-up for contact-rich tasks. This includes considering two research questions: 1) For skill transfer - how can a contact-rich skill be quickly transferred to a different robot (e.g. transferring a bolt-nut separation skill from a high-precision robot to a low-precision robot)? 2) For skill augmentation - how can existing contact-rich skills be used to create new contact-rich skills (e.g. augmentation of rigid-material skills to deal with soft materials)?The project will develop a portfolio of research into the science of digital skills for contact-rich tasks, focusing on common manufacturing tasks such as bolt-nut assembly/disassembly, peg-hole insertion/separation, and shaft-ring assembly/disassembly. The ability to transfer and augment digital skills for contact-rich tasks will allow automation systems to be implemented on a larger scale, with minimal manual setting and fine-tuning required. STAMAN aims to create transferrable and augmentable digital skills that will underpin the development of mass machine skills for future manufacturing, similar to how industrial robots have contributed to modern mass production.The proposed research encourages more use of robots in assembly (e.g. automotive, aerospace, electronics, etc.) and disassembly (e.g. repairs, remanufacturing and recycling), and thus directly contributes to the UK's Made Smarter initiative and the circular economy goals.

制造中的许多装配和拆卸任务具有小间隙和有限的可及性，例如轴孔插入/分离和螺栓螺母装配/拆卸。在这些接触丰富的任务中使用机器人比没有物理接触(例如计算机视觉检查)或简单接触(例如切割、焊接、拾取和放置)的任务要复杂得多。到目前为止，机器人在接触丰富的任务中的部署一直有限。接触丰富的任务涉及复杂的形状、小间隙或可变形材料，由于可能发生卡住和楔入事件，因此对机器人来说尤其具有挑战性。我们之前的研究已经调查了允许机器人使用两种主要的基于人工智能的途径学习接触丰富的技能(例如复杂的运动计划和力控制策略)的技术：(1)从试错中自我学习，(2)从人类演示中学习。伯明翰和谢菲尔德两所参与研究的大学分别拥有(1)和(2)方面的研究经验。当前研究中观察到的一个关键挑战是，在许多情况下，机器人的丰富接触技能无法由其他具有不同运动特性(如精确度、精确度和刚性)的机器人执行，或者无法应用于具有不同变化的新任务(例如，物体几何形状和材料的差异)。这是因为机器人接触丰富的技能，即控制策略和运动计划，通常是针对特定任务获得的，不能被新机器人或新任务采用。STAMAN的愿景是创建基于人工智能的机制，允许机器人共享和重新创建获得的数字技能(例如，运动和力/力矩控制策略)，以便轻松地自动化扩展接触丰富的任务。这包括考虑两个研究问题：1)对于技能转移-如何将接触丰富的技能快速转移到不同的机器人上(例如，将螺栓螺母分离技能从高精度机器人转移到低精度机器人)？2)对于技能增强-如何使用现有的丰富接触技能来创造新的接触丰富技能(例如，增强刚性材料技能以处理软材料)？该项目将开发一系列针对接触丰富任务的数字技能科学的研究，重点是常见的制造任务，如螺栓螺母组装/拆卸，销孔插入/分离、轴环装配/拆卸。为联系人丰富的任务转移和增强数字技能的能力将使自动化系统能够在更大范围内实施，所需的手动设置和微调最少。斯塔曼的目标是创造可转移和可扩展的数字技能，这些技能将为未来制造的大规模机器技能的发展奠定基础，类似于工业机器人对现代大规模生产的贡献。拟议的研究鼓励在装配(例如汽车、航空航天、电子产品等)中更多地使用机器人。此外，该项目还将用于制造和拆卸(如维修、再制造和回收)，从而直接促进英国的Made Smarter倡议和循环经济目标。