Bipedal Robot

在此建议下，我们寻求获得最先进的双足机器人来研究动态腿部运动。我们的总体科学目标是提高当今有腿机器人在速度、效率和多功能性方面的性能。在这种情况下，我们的研究是专门针对如何利用这种机器人的机械动力学，以提高其性能的问题。也就是说，我们试图理解一个有腿的机器人如何从重力、惯性和弹性振荡驱动的运动中受益，以减少使用的能量，提高峰值功率输出，或帮助稳定。我们也对不同步态的使用感兴趣，比如走路或跑步。这些目标受到生物系统的启发：人类和动物在运动方面表现出非凡的表现，他们通过巧妙地利用身体的动力学来做到这一点。我们试图在机器人系统中实现这些能力，首先使用概念模型，模拟和基于模型的优化来理解驱动自然机械动力学开发的基本物理原理。例如，我们已经能够证明，如何使用不同的步态可以大大减少能量消耗，提高未来有腿机器人的运动速度。结合这些基本原理，我们开发了控制方法，在真实的机器人上实现相同的增益。这是一个重要的步骤，因为速度、效率和多功能性等性能标准只有在实际的硬件实现中才能完全体现出来。为此，我们寻求获得一个商业级，双足机器人研究平台，能够自主动态腿运动。也就是说，它的能量重量比使得奔跑和跳跃成为可能。机器人的臀部、膝盖和脚踝都有完全驱动的关节。这些关节是扭矩驱动的，以便于控制，而执行器只表现出很小的反射惯性，以强调机械结构的自然动力学。机载传感器可以监控所有关节，通过惯性测量单元估计机器人的姿态，并在机器人前方的地面进行三维测量。传感器和执行器可以在开放访问的低级应用程序编程接口中直接访问，用于开发自定义控制算法。该机器人与实验环境组成一个单元，实验环境包括一个计算机控制的仪器跑步机，可以记录地面反作用力，一个视觉运动捕捉系统和一个架空龙门。这种环境将允许安全有效的实验，仪器将提供机器人状态的地面真实测量。这些对于校准和验证目的是必要的，并且可以进一步用于控制。测量的运动学和动力学也是不同步态的主要特征，是我们工作的核心焦点。

With this proposal, we seek to acquire a state-of-the-art bipedal robot to study dynamic legged locomotion. Our overall scientific goal is to improve the performance of today’s legged robots with respect to their speed, efficiency, and versatility. In this context, our research is specifically targeted at the question of how to exploit the mechanical dynamics of such a robot to improve its performance. That is, we seek to understand how a legged robot can benefit from motions driven by gravity, inertia, and elastic oscillations to reduce the amount of energy used, to improve peak power output, or to aid with stability. We are also interested in the usage of different gaits, such as walking or running.These goals are inspired by biological systems: humans and animals show remarkable performance with regard to locomotion and they do so by cleverly exploiting the dynamics of their body. We seek to enable these abilities in robotic systems, by first using conceptual models, simulations, and model-based optimizations to understand the fundamental physical principles that drive the exploitation of mechanical dynamics in nature. For example, we have already been able to demonstrate, how using different gaits can substantially reduce energy consumption and increase locomotion speed of future legged robots. Incorporating these fundamental principles, we then develop control approaches to realize the same gains on real robots. This is an important step, as performance criteria such as speed, efficiency, and versatility only fully manifest themselves in actual hardware implementations.To this end, we seek to acquire a commercial-grade, bipedal robotic research platform that is capable of autonomous dynamic legged locomotion. That is, its power-to-weight ratio is such that running and hopping gaits are possible. The robot has fully actuated joints at its hip, knees, and ankles. These joints are torque driven to facilitate control, while actuators only exhibit a small reflected inertia to emphasize the natural dynamics of the mechanical structure. On-board sensors allow monitoring all joints, estimating the posture of the robot via an inertial measurement unit, and surveying the ground in front of the robot in three dimensions. Sensors and actuators are directly accessible within an open-access low-level application programming interface for the development of custom control algorithms. The robot forms a unit with an experimental environment consisting of a computer-controlled instrumented treadmill which can record ground reaction forces, a visual motion capture system, and an overhead gantry. This environment will allow for safe and effective experimentation and the instruments will deliver ground truth measurements of the state of the robot. These are imperative for calibration and validation purposes and can be further used for control. The measured kinematics and kinetics are also a main characteristic of different gaits, a core focus of our work.