EAGER: CPR-COVID-19 Prevention Robot in Dense Areas

受COVID-19疫情影响，该项目将开发一款机器人，以了解公共场所或办公室的行人是否遵守社交距离指引。该项目将开发利用机器学习、计算机视觉和机器人运动规划的新方法，以确定行人在有限区域内移动时的位置。实时了解行人的移动可以帮助社交距离的努力，最大限度地减少COVID-19的传播，并可以更广泛地增强人机交互。潜在的挑战包括开发新的导航算法，可以计算机器人在中等和高密度人群中的无碰撞路径。行人之间的导航将被公式化为部分可观察的马尔可夫决策过程，并使用深度强化学习来解决，特别是关注邻近策略优化。行人跟踪方法将基于一个新的概念，即正面倒易速度障碍，它使用每个行人运动的椭圆近似，并通过考虑中间目标和避免碰撞来估计潜在的动态。计划中的方法还将能够通过以智能方式移动机器人来处理行人之间的遮挡，以改善它从传感器接收到的信息。该项目将使用商品传感器，包括摄像头和2D激光雷达，以了解行人的运动，并检查社交距离的限制。最后，该项目将研究使用机器人影响行人行为的技术。该奖项反映了NSF的法定使命，并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，被认为值得支持。

Motivated by the COVID-19 pandemic, this project will develop a robot to understand whether pedestrians in public places or offices are maintaining social distancing guidelines. The project will develop new methods that leverage machine learning, computer vision, and robot motion planning to ascertain the positions of pedestrians as they move in a confined area. Real-time understanding of pedestrian movements can assist social distancing efforts, minimizing the spread of COVID-19, and can more broadly enhance human-robot interactions.The underlying challenges include development of new navigation algorithms that can compute collision-free paths for a robot in medium and high-density crowds. Navigation among pedestrians will be formulated as a Partially-Observable Markov Decision Process and solved using deep reinforcement learning, particularly focusing on Proximal Policy Optimization. The pedestrian-tracking approach will be based on a novel concept of Frontal Reciprocal Velocity Obstacles, which uses an elliptical approximation of each pedestrian motion and estimates the underlying dynamics by considering intermediate goals and collision avoidance. The planned approach will also be able to handle occlusions among pedestrians by moving the robot in an intelligent way to improve the information that it receives from its sensors. The project will use commodity sensors, including cameras and 2D LIDARs, to understand pedestrian movements and check for social distance constraints. Finally, this project will investigate techniques to influence pedestrian behavior using robots.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.