HUME: A Bipedal Robot for Human-Centered Hyper-Agility

Our broad goals are to understand the physical capabilities and physiology of human movement for use in the design of machines with similar abilities. Here, we discuss the development of a hyper agile bipedal robot, HUME, capable of very quickly traversing rough terrains that are at the extrema of what humans can overcome on two feet. We will refer to this skill as Human-Centered HyperAgility (HCHA). In particular the extrema of HCHA includes freerunning-like capabilities on vertical surfaces (Parkour). We aim to be the first to endow true HCHA to a human-sized bipedal robot through the following approach: (1) develop a simulation environment for dynamic single and multi-contact maneuvers in highly irregular terrains (HCHA), (2) from this simulation environment extract the performance parameters associated with HCHA, (3) design an actuator capable of delivering the design specifications, (4) build a humansized bipedal robot capable of delivering the desired performance, (5) build a boom system and an irregular terrain to support the robot, provide the environment and capture precision data of the free-body dynamics. We believe true HCHA is a very important capability because of its direct impact in the design of human assistive devices for all terrains and the design of next generation semi-autonomous bipedal robot. Laboratories around the world have produced outstanding designs covering many different areas of legged capabilities. For the purpose of analysis we consider robots’ performances as a function of speed, agility and efficiency, where speed is the center of mass absolute velocity, agility is the complexity of traversable terrain and efficiency is the cost of transport. The reader can imagine the robots reviewed below as points in 3-D space with speed, agility and efficiency as the principle performance axes, where our target performance (HCHA) would maximize speed and agility while attempting to optimize efficiency. The first biped robot to achieve “quasi-dynamic” walking was described in [13] followed by dynamically stable robots discussed in [17]. Passive dynamic walking was introduced in [15], [16] and later, in [4], [3]. Various bipedal humanoids [6], [11], [12], [14] are skilled at mimicking humans in a variety of terrains and speeds, but as terrains become more complex their speeds decrease from those of the human. Bipedal robots built for high speeds [22], [2], [29] have all shown that they can cover level terrain quickly but have not been used to traverse very rough environments at these speeds. Series elastic actuation (SEA) [18], [23] has spurred the design of various compliant robots [20], [19], [10], and has been extended in the form of actuators with mechanically adjustable compliance in [8], [5], [9], [7]. Although the following are quadrupeds, [21], [1] are at the frontiers of hyper agility. Both have the ability to overcome rough terrains quickly and therefore it is one of our objective to endow these capabilities in a bipedal robot. Our work is a departure from previous designs in that we seek to reach the extrema of human movement in cluttered environments. This entails a clear understanding of hyper-agility in these environments and a strong competence in building actuators and humanoid robot mechatronics. UT-Austin has teamed up with Meka Robotics to perform the study for and delivery of a machine that can show significant advancements toward HCHA. We are aware of the new PETMAN robot by Boston Dynamics which delivers high mechanical power and speed but its detailed architecture is uncertain to us. The effort for achieving HCHA can obviously not be centered only around hardware design but needs to be complemented by stability planners and compliant controllers. Our efforts in trajectory planning in irregular and extreme terrains has been strong recently, with publications in [24], [25], [28]. More recently we have conducted experiments on using whole body compliant control on a mobile humanoid robot that balances discussed in our submission [27].