NRI: FND: Immersive whole-body teleoperation of wheeled humanoid robots for dynamic mobil manipulation

建筑和物流行业的工人需要体力劳动，还有一些危险的工作，如救灾、消防、警察、伐木和采矿。在美国，这些工人中有许多因工作场所事故而受伤、致残或死亡。虽然机器人有潜力保护这些工人的福祉，但它们还不具备执行这些活动所固有的高要求体力任务的能力。一个关键缺失的方面是直觉地协调他们的身体来放大他们施加在物体上的力的智能，比如当人类自然地靠在沉重的门上打开它时。这个项目的重点是开发将人类的控制智能与机器的体力和耐力相结合的技术。在该框架中，可穿戴设备测量人类操作员整个身体的姿势，使得远程类人机器人可以同时再现相同的全身位置。这个项目的新颖之处在于使人类操作员能够同时感受到机器人对其周围环境施加的力。例如，一名消防员通过自己的动作来命令机器人推动重物，同时感受机器人是否用力不足，物体没有移动，或者是否用力过度而失去平衡。这项研究的影响是通过使人类工人远离危险来保护他们的福祉，同时利用他们的领域专业知识和运动智能来执行体力劳动。这项研究促进了本科生和代表性不足的少数民族在研究中的纳入，与K-12学生建立了外展活动，并在会议和大学活动期间提出了该系统的演示。本项目探索一种新的双边远程操作框架，使人类操作员能够控制远程轮式人形机器人执行动态移动操作（DMM）。DMM是结合了有力的操作和灵活的运动的体力任务。本研究的关键假设是，通过全身触觉装置的双侧遥操作可以在非结构化的野外环境中实现安全、直观的DMM。为了验证这一假设，将在三年内实现三个具体目标。目标1：实现实现本研究的硬件，包括(i)一个原型全身触觉设备，它可以读取操作员的姿势来控制远程机器人，并从机器人与环境的交互中提供丰富的运动、视觉和听觉反馈，以及（ii）名为SATYRR的机器人，它结合了腿系统、轮式运动和拟人化上半身的优点，可以进行双臂操作。目的2：研究DMM的全身双侧遥操作策略。设计一个直观的人机运动映射，高带宽控制，并实施力反馈方案，即使在大冲击下也能向操作员提供可解释的反馈。目标3：利用共享自主权提高DMM的安全性。设计模型预测控制方案，尽可能紧跟用户输入，同时防止操作员输入损坏机器人。为了评估计划的框架，将建立一个现实的实验室任务环境，其中包括推动大型物体，打开阻塞的门，在外部干扰下保持平衡，以及携带和投掷有效载荷。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

Physically demanding labor is required of workers in construction and logistics, as well as dangerous jobs such as disaster relief, firefighting, policework, logging, and mining. Many of these workers in the US are injured, disabled, or die due to workplace incidents. Although robots have the potential to protect the well-being of these workers, they do not yet possess the capability to perform the demanding physical tasks which are inherent to these activities. A key missing aspect is the intelligence to intuitively coordinate their body to amplify the forces that they apply to objects, such as when humans naturally lean against a heavy door to open it. This project focuses on developing the technology to combine the control intelligence of humans with the physical strength and endurance of machines. In this framework, a wearable device measures the posture of the entire body of a human operator such that the remote human-like robot may reproduce the same whole-body position at the same time. The novelty of this project is to enable the human operator to simultaneously feel the forces that the robot applies to its surroundings. For instance, a firefighter commands the robot to push heavy debris by performing the motion herself, while simultaneously feeling whether the robot is exerting insufficient force and the object isn’t moving, or whether it is exerting excessive effort and losing balance. The impact of this research is the protection of the well-being of human workers by taking them away from danger, while leveraging their domain expertise and their motor intelligence to perform physically demanding labor. This research promotes the inclusion of undergraduate and underrepresented minorities in research, creates outreach activities with K-12 students, and proposes demonstrations of the system during conferences and University events.This project explores a novel bilateral teleoperation framework which enables a human operator to control a remote wheeled humanoid robot to perform Dynamic Mobile Manipulation (DMM). DMM are physical tasks which combine forceful manipulation and agile locomotion. The key hypothesis in this proposal is that bilateral teleoperation via a whole-body haptic device can achieve safe and intuitive DMM in unstructured field environments. To validate this hypothesis, three specific aims will be pursued over three years. Aim 1: Implement the hardware that enables this research, composed of (i) a prototype whole-body haptic device which reads the operator’s posture to control the remote robot, and provides rich kinesthetic, visual, and audial feedback from the robot’s interaction with its environment, and (ii) the robot named SATYRR, which combines advantages of legged systems, wheeled locomotion, and an anthropomorphic upper-body for dual-arm manipulation. Aim 2: Study whole-body bilateral teleoperation strategies for DMM. Devise a human-to-robot motion mapping for intuitive, high-bandwidth control, and implement force feedback schemes that deliver interpretable feedback to the operator even under large impacts. And Aim 3: Improve DMM safety using shared autonomy. Devise model predictive control schemes to follow the user's input as closely as possible while preventing operator inputs that would damage the robot. To evaluate the planned framework, a realistic laboratory task environment will be built with an obstacle course that includes pushing large objects, opening blocked doors, balancing under external disturbances, and carrying and throwing a payload.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.