Robust, Cooperative and Intelligent Dexterous Manipulation in Complex Environments

在过去的几十年里，机器人灵巧操作越来越受到人们的关注，其应用范围广泛，例如制造业的智能工厂，军事应用，仓库，手术和假肢以及医疗保健服务。尽管机器人、人工智能和计算技术取得了重大进展，但机器人执行一般任务的能力与人类相比仍有很大差距。拟议的工作解决了机器人和自动化领域公认的主要挑战之一。拟议研究的结果将使与人类互动的机器人系统更可靠，适应性更强，操作更高效。合作操作的研究将有助于该行业提供操作便利，提高经济效益，并使需要医疗干预和身体帮助的人受益。 短期目标是实现：1）单臂的触觉和顺应性操纵，以提高装备有抓握手和触觉传感的操纵器的灵活性，用于制造和健康护理应用; 2）用于复杂环境中的任务的灵巧的协作机器人操纵，其中需要双手和多个机器人操纵，例如拧紧，移除危险材料，手术，和装配任务; 3）智能人机灵巧协同操作，以提高人类的安全性，生产力，并促进人类指导的行为学习。 将通过广泛的实验研究来研究以下问题。将机械手与7自由度机械手集成，以提高机械手的可操作性和灵活性。顺应类人操纵力控制，新的协调控制自由运动，新的自适应导纳控制与环境相互作用时，将开发和比较基于学习的方法。针对复杂环境下多机器人的避碰协同操作问题，提出了基于力/运动协调的多机器人协同操作方案，并对其进行了优化和性能评价。将开发一个人机界面，以预测人类合作伙伴的目标，用于人机协作任务。有效的人机协同操作方法，多模态，稳定和鲁棒的人机灵巧操作和新的强化学习方法在复杂环境中的人机协同操作将被提出。该研究将为高素质人才提供跨学科的培训环境，这些人才具有更多的实践能力，为机器人，控制，机电一体化，信号处理和人工智能等关键技术领域的创新做出贡献。发展高科技自动化和机器人产业有可能为加拿大经济做出重大贡献，这项研究的成果将提高加拿大在工业自动化和智能系统方面的地位。

Over the past decades, increasing attention has been paid to robotic dexterous manipulation for a broad range of applications, such as smart factories in manufacturing, military applications, warehouses, surgery and prosthetics, and health care services. Despite the significant progress of robotics, artificial intelligence and computing technologies, the ability of robots to perform general tasks is still far from that of human beings. The proposed work addresses one of the recognized major challenges in robotics and automation. The results of the proposed research will enable more reliable, adaptable, and efficient operation of robotic systems interacting with humans. The research in cooperative manipulation will help the industry to offer operating convenience and increase economic efficiency as well as benefit to people requiring medical interventions and physical assistance. The short term objectives are to achieve: 1) tactile and compliant manipulation of a single arm to improve the dexterity for manipulators equipped with grasping hands and tactile sensing for manufacturing and health care applications; 2) dexterous cooperative robotic manipulation for tasks in a complex environment where bimanual and multiple robotic manipulations are required, such as screwing, removal of hazardous materials, surgery, and assembly tasks; 3) intelligent human-robot dexterous co-manipulation to improve human safety, productivity, and facilitate human-guided behavior learning. The following issues will be studied with extensive experimental studies. Robotic hands will be integrated with 7-DOF manipulators to improve the manipulability and the dexterity. Compliant human-like manipulations with force control, new coordination control for free motion, novel adaptive admittance control when interacting with the environment, and comparisons with learning-based approach will be developed. Cooperative manipulation between multiple manipulators with collision avoidance in a complex environment, force/motion coordination schemes on cooperative tasks will be developed, optimized and evaluated with proper performance indices. A human-machine interface to anticipate the objective of the human partner for human-robot collaboration tasks will be developed. Effective human-robot co-manipulation approaches, multi-modal, stable and robust human-robot dexterous manipulation and new reinforcement learning approach for human-robot co-manipulation in complex environments will be proposed. The research will provide an interdisciplinary training environment to highly qualified personnel with more hands-on ability to contribute to innovation in the key technological areas in robotics, control, mechatronics, signal processing and AI. Developing the high-technology automation and robotic industry has the potential to contribute significantly to the Canadian economy and the outcomes of this research will enhance Canada's profile in industry automation and intelligent systems.