AIARA: Artificial Intelligence Enabled Highly Adaptive Robots for Aerospace Industry

通过使用自适应机器人的敏捷制造是提供先进制造，工业4.0，以实现更高效，精益和成本效益的定制，大规模但小批量的产品。对于严重依赖体力劳动的制造业来说，这被认为是最终的解决方案。航空航天行业是严重缺乏自动化的行业之一，这导致新飞机订单积压，并阻碍其采取足够快的行动来采用更高效的飞机设计和先进材料。通过在制造新飞机时使用机器人来提高自动化水平，不仅迫切需要降低成本，而且还需要提高航空航天行业的质量和安全。然而，在汽车工业和电子设备的装配线上使用的传统工业机器人，由于批次小、零部件大、产品多样化以及操作的高度复杂性和多样性，已经不能满足航空航天行业的需要。因此，在航空航天行业，目前为每一项特定任务编程或教授机器人的做法即使不是徒劳的，也是有限的。在先进制造业和工业4.0中，机器人是智能的，适应性很强，可以被训练来处理不同的设备、工具、产品和材料，而不需要显式编程。然而，基于人工智能的学习方法需要大量数据来捕获所有可能的物理体验来训练机器人，这可能过于昂贵或无法获得。机器人学的最新进展表明，从合成的机器人经验和模拟中学习是可行的。在拟议的项目中，我们的目标是开发一种方法，利用来自仿真和虚拟环境的学习结果来稳健地训练适用于广泛航空航天制造过程的多臂自适应机器人。这一研究伙伴关系将加拿大高科技领域的著名公司UBC、Kinova Inc.和Element AI聚集在一起，并与德国航空航天中心DLR合作，为航空航天行业更有效的制造构建解决方案。

Agile manufacturing by the use of adaptive robots is the provision of advanced manufacturing, Industry 4.0, to enable more efficient, lean and cost-effective production of customized, large-scale but small batch size products. It is considered to be the ultimate solution for manufacturing industries hindered by their heavy reliance on manual labor. Aerospace industry is one of those industries that suffers greatly from lack of automation causing a backlog of new aircraft orders and prevents it from moving fast enough to adopt more efficient aircraft designs and advanced materials. An increased level of automation via the use of robots in manufacturing new aircrafts is pressing for not only cost reduction but also improved quality and safety in the aerospace industry. However, traditional industrial robots used in assembly lines of automotive industry and electronic devices is inadequate for the aerospace industry, because of small batch sizes, large components, diversity of products and a high level of complexity and variation in operations. Thus, the current practice of programming or teaching a robot for every specific task is limited, if not futile, in the aerospace industry. In advanced manufacturing and Industry 4.0, robots are intelligent, highly adaptive and can be trained to handle different equipment, tools, products and materials without a need for explicit programming. However, artificial intelligence-based learning methods require a large volume of data for capturing all possible physical experiences to train the robot, which can be too expensive or unavailable. Recent advances in robotics demonstrate the feasibility of learning from synthetic robot experiences and simulations. In the proposed project, we aim to develop a methodology to use learning results from simulation and virtual environments to robustly train a multi-arm adaptive robot for a wide range of aerospace manufacturing processes. This research partnership brings together the UBC, Kinova Inc., and Element AI, both prominent Canadian companies in the hi-tech sector, and in collaboration with the German Aerospace Centre DLR to build solutions for more effective manufacturing in aerospace industry.