NRI: Human Cognition Assisted Control of Industrial Robots for Manufacturing

以工业机器人为主导的先进制造业在美国经济中发挥着越来越大的作用。机器人被用来进行装配、焊接、材料处理和制造。尽管这种互动在制造业的每个阶段都变得越来越普遍，但机器和人之间完美的共生关系仍然很遥远。正因为如此，大多数机器人在制造业中的应用目前仅限于对技能要求相对较低的领域。这限制了机器人在增强人类操作者、提高生产力和生活质量方面的全部潜力。随着认知神经科学和脑接口技术的最新进展，将人类认知思维过程直接连接到机器人和机器是可能的，从而直接控制现实世界的应用。通过使用传感器收集大脑信号并分析思维过程，当人类采取特定行动或思考行动时，大脑中发生的许多活动可以通过快速计算识别并与已知信号相匹配。这种新的人机通信范式将通过开发三种制造场景来演示。该项目还将广泛适用于制造业以外领域的机器人系统设计，包括远程外科手术、康复和太空探索。这项多学科研究结合了制造业、计算机科学和机器人技术，其结果有可能提高未来制造工厂的生产率，并可能导致新的商业投资，这将有助于美国保持机器人和制造业的全球领导地位，扩大未被充分代表的群体参与研究，并对工程教育产生积极影响。新的人机通信系统的发展将面临重大的未来挑战，该系统允许操作员执行复杂的高技能任务。假设的范式将通过满足以下智力挑战来探索：(i)研究一种新的方法，通过使用称为“行为”的语法来想象简单的动作，从而向机器人传达运动命令；（ii）新的脑机模式和算法来对这些行为进行分类；（iii）基于意图的系统，该系统根据人类操作员计划完成的最有可能的事件序列自动完成机器人动作。三种机器人制造场景将探讨在制造环境中基于人类认知的交互：装配，直接控制和通过对象识别的质量控制。最后，通过使用非侵入性脑机接口，将展示机器人的广泛日常应用。

Advanced manufacturing, driven by industrial robots, is playing an increasing role in US economy. Robots are being used to carry out assembly, welding, material handling and fabrication. Even as such interactions are becoming more common in every phase of manufacturing, a perfect symbiotic relationship between machines and human beings is still very far away. Because of this, a majority of the robotic applications in manufacturing are currently limited to areas where a relatively low level of skill is required. This has restricted the full potential of robotics to augment human operators and improve productivity and quality of life. With recent advances in cognitive neuroscience and brain interface technologies, connecting the human cognitive thought process directly to robots and machines is possible, resulting in direct control of real world applications. By collecting the brain signals using sensors and analyzing the thought processes, many activities that take place inside the brain when humans take specific actions or think of actions can be identified and matched to known signals using fast computation. This new human-robot communication paradigm will be demonstrated by developing three manufacturing scenarios. The project will also have broad applicability in the design of robotic systems in fields outside manufacturing, including telesurgery, rehabilitation and space exploration. Results from this multidisciplinary research, which combines manufacturing, computer science and robotics, have the potential to improve the productivity of future manufacturing plants and can lead to new commercial ventures, which will help the US maintain global leadership in robotics and manufacturing, broaden participation of underrepresented groups in research, and positively impact engineering education.Significant future challenges in the development of a new human-robot communication system, which allows operators to perform complex high skilled tasks, will be addressed. The postulated paradigm will be explored by meeting the following intellectual challenges: (i) researching a novel methodology for communicating motion commands to a robot by imagining simple actions using a grammar called "actemes," (ii) new brain-computer mode and algorithms to classify these actemes and, (iii) an intent-based system that auto-completes robotic actions based on most likely sequence of events that human operators are planning to complete. Three robotic manufacturing scenarios will be explored to demonstrate the human cognition based interactions in manufacturing environment: assembly, direct control, and quality control through object recognition. Finally, by using a non-invasive brain-computer interface a wide range of day-to-day applications of robotics will be demonstrated.