Realizing dynamic and efficient bipedal locomotion on the humanoid robot DURUS

Abstract

This paper presents the methodology used to achieve efficient and dynamic walking behaviors on the prototype humanoid robotics platform, DURUS. As a means of providing a hardware platform capable of these behaviors, the design of DURUS combines highly efficient electromechanical components with "control in the loop" design of the leg morphology. Utilizing the final design of DURUS, a formal framework for the generation of dynamic walking gaits which maximizes efficiency by exploiting the full body dynamics of the robot, including the interplay between the passive and active elements, is developed. The gaits generated through this methodology form the basis of the control implementation experimentally realized on DURUS; in particular, the trajectories generated through the formal framework yield a feedforward control input which is modulated by feedback in the form of regulators that compensate for discrepancies between the model and physical system. The end result of the unified approach to control-informed mechanical design, formal gait design and regulator-based feedback control implementation is efficient and dynamic locomotion on the humanoid robot DU-RUS. In particular, DURUS was able to demonstrate dynamic locomotion at the DRC Finals Endurance Test, walking for just under five hours in a single day, traveling 3.9 km with a mean cost of transport of 1.61-the lowest reported cost of transport achieved on a bipedal humanoid robot.

Dynamic Walking with Compliance on a Cassie Bipedal Robot

“…We argue that a large portion of this particular trade-off can be compensated through the inclusion of the compliance which exists in in the physical model. Other work considering prismatic passive compliant elements [20], and an embedded complaint controller [24], have also demonstrated these concepts on hardware. To demonstrate this, we controlled the full model walking with a reference trajectory designed based on the simple model.…”

Rapid Trajectory optimization Using C-FROST with Illustration on a Cassie-Series Dynamic Walking Biped

“…The ability to run many different gait optimization in parallel is particularly useful for learning-based control approaches as described in [7], [8], and can also be used in the mechanical design process of a robot, where instead of gait characteristics varying, link lengths, mass distributions, and gear ratios are varied [27], [32].…”

“…The implementation of C-FROST also makes it easier to deploy parallel computation on cloud servers for the generation of a large set of different gaits, a feature that allows the gait library approach presented in [7] to be practical on highdimensional 3D bipedal robots and advanced exoskeletons [14]. Beyond generating families of gaits for various walking speeds, directions, and terrain slopes, the ability to generate gaits in parallel for a family of models is also useful for iterative design of legged mechanisms [6], [27].…”

Realizing Learned Quadruped Locomotion Behaviors through Kinematic Motion Primitives

“…However, it is important to note that the controller learned is modelfree, but the training is normally done in simulation which is often not realistic. This reality gap is also a well known problem faced by traditional model based optimization tools for walking [18], [19], [20]. One of the main goals of this paper is to approach this problem of reality gap by using the tools used in formal methods.…”

“…Formal methods usually employ the optimization tools only for generating the reference gaits (trajectories), and then these trajectories are tracked at the joint level by model free methods (e.g. PD control) [18], [21], [22]. Therefore, we would like to view D-RL based training as a trajectory generator (optimization tool) more than the onboard compute platform for walking.…”