17 March 2021 in IEEE Robotics and Automation Letters
In this letter, we describe an approach to achieve dynamic legged locomotion on physical robots which combines existing methods for control with reinforcement learning. Specifically, our goal is a control hierarchy in which highest-level behaviors are planned through reduced-order models, which describe the fundamental physics of legged locomotion, and lower level controllers utilize a learned policy that can bridge the gap between the idealized, simple model and the complex, full order robot. The high-level planner can use a model of the environment and be task specific, while the low-level learned controller can execute a wide range of motions so that it applies to many different tasks. In this letter, we describe this learned dynamic walking controller and show that a range of walking motions from reduced-order models can be used as the command and primary training signal for learned policies. The resulting policies do not attempt to naively track the motion (as a traditional trajectory tracking controller would) but instead balance immediate motion tracking with long term stability. The resulting controller is demonstrated on a human scale, unconstrained, untethered bipedal robot at speeds up to 1.2 m/s. This letter builds the foundation of a generic, dynamic learned walking controller that can be applied to many different tasks.
28 January 2020 in IEEE Robotics and Automation Letters
Frictional anisotropy, as produced by the directionality of scales in snake skin, is necessary to propel snakes across flat, hard surfaces. This work illustrates the design, fabrication, and testing of a snake-inspired skin based on kirigami techniques that, when attached to a soft snake robot, improves the robot's locomotion capabilities when implementing a lateral undulation gait. Examination of snake scales in nature informed the shape and texture of the synthetic scales, which are activated through the buckling of kirigami lattices. Biological snakes have microornamentation on their scales, which is replicated by scoring ridges into the plastic skin. This microornamentation contributes to the lateral resistance necessary for lateral undulation. The skin's frictional properties were experimentally determined, as were their contributions to the locomotion of the robot across a flat, hard, textured surface. Contributions to locomotion from scale profile geometry, scale microornamentation, and scale angle of attack were identified. The range of longitudinal COF ratios was 1.0 to 3.0 and the range of lateral COF ratios was 0.9 to 3.3. The highest performing skin was the triangular scale profile with microornamentation, producing a velocity of 6 mm/s (0.03 BL/s) which is an increase of 335% over the robot with no skin when activated to maximum achievable curvature.
9 July 2019 in IEEE Transactions on Robotics
In this paper, we present a set of geometric principles for understanding and optimizing the gaits of drag-dominated kinematic locomoting systems. For systems with two shape variables, the dynamics of gait optimization are analogous to the process by which internal pressure and surface tension combine to produce the shape and size of a soap bubble. The internal pressure on the gait curve is provided by the flux of the curvature of the system constraints passing through the surface bounded by the gait, and surface tension is provided by the cost associated with executing the gait, which when executed at optimal (constant-power) pacing is proportional to its pathlength measured under a Riemannian metric. We extend these principles to work on systems with three and then more than three shape variables. We demonstrate these principles on a variety of system geometries (including Purcell’s swimmer) and for optimization criteria that include maximizing displacement and efficiency of motion for both translation and turning motions. We also demonstrate how these principles can be used to simultaneously optimize a system’s gait kinematics and physical design.
13 May 2022 in IEEE Transactions on Robotics
Isolated mechanical systems—e.g., those floating in space, in free-fall, or on a frictionless surface—are able to achieve net rotation by cyclically changing their shape, even if they have no net angular momentum. Similarly, swimmers immersed in “perfect fluids” are able to use cyclic shape changes to both translate and rotate even if the swimmer-fluid system has no net linear or angular momentum. Finally, systems fully constrained by direct nonholonomic constraints (e.g., passive wheels) can push against these constraints to move through the world. Previous work has demonstrated that the displacement induced by these shape changes corresponds to the amount of constraint curvature that the gaits enclose. Properly assessing or optimizing the utility of a gait also requires considering the time or resources required to execute it: A gait that produces a small displacement per cycle, but that can be executed in a short time, may produce a faster average velocity than a gait that produces more displacement, but takes longer to complete a cycle at the same instantaneous effort. In this paper, we consider gaits under two instantaneous measures of effort. For each of these costs, we demonstrate that fixing the average instantaneous cost to a unit value allows us to transform the effort costs into time-to-execute costs for any given gait cycle. We then illustrate how the interaction between the constraint curvature and these costs leads to characteristic geometries for optimal cycles, in which the gait trajectories resemble elastic hoops distended from within by internal pressures.
9 March 2021 in Physical Review E
In this paper we identify optimal swimming strategies for drag-dominated swimmers with a passive elastic joint. We use resistive force theory to obtain the dynamics of the system. We then use frequency-domain analysis to relate the motion of the passive joint to the motion of the actuated joint. We couple this analysis with elements of the geometric framework introduced in our previous work aimed at identifying useful gaits for systems in drag-dominated environments to identify speed-maximizing and efficiency-maximizing gaits for drag-dominated swimmers with a passive elastic joint.
24 January 2022 in Bioinspiration & Biomimetics
Modeling system dynamics becomes challenging when the properties of individual system components cannot be directly measured, and often requires identification of properties from observed motion. In this paper, we show that systems whose movement is highly dissipative have features which provide an opportunity to more easily identify models and more quickly optimize motions than would be possible with general techniques. Geometric mechanics provides means for reduction of the dynamics by environmental homogeneity, while the dissipative nature minimizes the role of second order (inertial) features in the dynamics. Here we extend the tools of geometric system identification to ‘shape-underactuated dissipative systems (SUDS)’—systems whose motions are more dissipative than inertial, but whose actuation is restricted to a subset of the body shape coordinates. Many animal motions are SUDS, including micro-swimmers such as nematodes and flagellated bacteria, and granular locomotors such as snakes and lizards. Many soft robots are also SUDS, particularly robots that incorporate highly damped series elastic actuators to reduce the rigidity of their interactions with their environments during locomotion and manipulation. We motivate the use of SUDS models, and validate their ability to predict motion of a variety of simulated viscous swimming platforms. For a large class of SUDS, we show how the shape velocity actuation inputs can be directly converted into torque inputs, suggesting that systems with soft pneumatic or dielectric elastomer actuators can be modeled with the tools presented. Based on fundamental assumptions in the physics, we show how our model complexity scales linearly with the number of passive shape coordinates. This scaling offers a large reduction on the number of trials needed to identify the system model from experimental data, and may reduce overfitting. The sample efficiency of our method suggests its use in modeling, control, and optimization in robotics, and as a tool for the study of organismal motion in friction dominated regimes.