SBIR Phase I: Modular and Updatable Artificial Intelligence (AI) for Robotics

这个小型企业创新研究（SBIR）第一阶段项目的更广泛的影响/商业潜力是为机器人行业的计算机视觉提供一种新颖的识别架构。该项目旨在使计算机学习没有排练，允许纠正的细节，目前在真实的世界环境。该项目的目标是提供一种解决方案，供计算机视觉客户使用，以立即解决他们的问题（无需将数据发送回重新训练整个网络），减少机器和客户停机时间和中断，同时提高生产力。最初的重点是具有计算机视觉限制的机器人技术，尽管该技术可能对其他行业有用。成功地改善基于计算机视觉的学习可以促进灾害响应，增强当前的身体能力，并使探索超越地球的边界。这个小型企业创新研究（SBIR）第一阶段项目将帮助创建一个框架，以克服排练要求，限制自动化机器人在逼真的动态环境中的效用。与人类相比，人工智能（AI）在快速积累知识而不忘记以前学到的知识方面仍然不灵活。使用AI的机器人目前仅用于非常有限且控制非常严格的环境中。在机器人的工作环境中可能发生的一切都必须包括在他们的训练集中。所提出的AI解决方案适合在动态环境中学习，而无需排练，同时在遇到信息时保持可扩展性。这项技术可以让机器人在他们的环境中接受训练。这个项目可以使视觉能力，导致灵活的学习示范，而无需排练在动态的机器人环境。这个奖项反映了NSF的法定使命，并已被认为是值得通过使用基金会的智力价值和更广泛的影响审查标准进行评估的支持。

The broader impact/commercial potential of this Small Business Innovation Research (SBIR) Phase I project is to provide a novel recognition architecture to computer vision in the robotics industry. The project seeks to enable computer learn without rehearsal, allowing corrections for details that are present in the real world environment. The aim of this project is a solution to be used by computer vision customers to solve their problems immediately (without sending data back to retrain the whole network), reducing machine and customer downtime and disruption while increasing productivity. The initial focus is on robotics with computer vision limitations though the technology may be useful to other industries. Success in improving computer vision-based learning could facilitate disaster responses, augment current physical abilities, and enable exploration beyond the boundaries of Earth. This Small Business Innovation Research (SBIR) Phase I project will help create a framework to overcome rehearsal requirements that limit automated robots’ utility within life-like, dynamic environments. Artificial intelligence (AI) remains inflexible compared to humans at quickly accumulating knowledge without forgetting what they have previously learned. Robots using AI are currently only used in environments that are very limited and are very tightly controlled. Everything that might happen in the robot’s work environment must be included their training set. The proposed AI solution is suited for learning in dynamic environments without rehearsal while maintaining scalability as information is encountered. This technology may allow robots to be trained within their environment. This project may enable visual capabilities leading to a demonstration of flexible learning without rehearsal within dynamic robotic environments.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.