Bulletin of the American Physical Society
APS March Meeting 2018
Volume 63, Number 1
Monday–Friday, March 5–9, 2018; Los Angeles, California
Session B50: Robophysics: Robotics Meets PhysicsFocus
|
Hide Abstracts |
Sponsoring Units: DBIO GSOFT Chair: Daniel Goldman, Georgia Institute of Technology Room: LACC 511B |
Monday, March 5, 2018 11:15AM - 11:51AM |
B50.00001: From Razor Clams to Robots: Drawing Engineering Inspiration from Natural Systems Invited Speaker: Anette Hosoi Many natural systems have evolved optimal strategies to perform certain tasks -- climbing, sensing, swimming -- within the limits set by the laws of physics. This observation can be used both to guide engineering design, and to gain insights into the form and function of biological systems. In this talk I will discuss both of these themes in the context of crawling snails, digging clams, and swimming microorganisms. We will discover how an analysis of the physical principles exploited by snails and clams leads to the development of novel robotic diggers and crawlers (RoboSnail and RoboClam) and explore the paddling strategies that result from the unique challenges faced by tiny swimmers. |
Monday, March 5, 2018 11:51AM - 12:03PM |
B50.00002: Validation of an Information Theoretic Measure of Locomotor Centralization Using Phase-Coupled Oscillator Models Amoolya Tirumalai, Izaak Neveln, Simon Sponberg Locomotor control differs in the degree of centralization: conceptually, the extent to which changes in local and global states are coupled. We recently developed a model-free mutual information-based centralization measure for animal and robotic systems. For this measure to be valid, it should reflect the centralization manifest in the network characteristics of phase-coupled oscillator models applied to locomotor data. We simulate and perturb a noisy six-oscillator system with preferred phase differences and network that produce the cockroach’s alternating tripod gait. We estimate the mutual information between perturbations and both single-oscillator and systemwide states (local and global responses). By varying the perturbations or networks we characterize the difference between local and global mutual information: our centralization measure. As network weighting increases, centralization increases irrespective of perturbation type. As network average clustering coefficient increases, centralization increases. This validation shows that our model-free measure predicts the behavior of a coupled oscillator model of cockroach locomotion, supporting general applicability of this centralization measure to living and engineered systems. |
Monday, March 5, 2018 12:03PM - 12:15PM |
B50.00003: Obstacle Modulation of Legged Robot Dynamics Feifei Qian, Zhichao Liu, Daniel Koditschek Legged robots have unique capabilities to negotiate obstacles. However, due to limited understanding of the complex interactions between robot legs and obstacles, many legged platforms circumnavigate obstacles to avoid large force disturbances. Our study aims to provide physical understanding of complex robot-obstacle interactions, enabling multilegged robots to exploit environment shape and generate desired dynamics by properly coordinating leg movements. With a previously characterized dependence of obstacle reaction force on contact position, we develop a predictive model that captures robot dynamics by considering obstacles as a superposed force field. We predict and empirically demonstrate the existence of asymptotically stable yaw angle equilibria in the horizontal plane dynamics of a quadruped robot running over periodically spaced “logs” (cylindrical obstacles). As obstacle spatial period is varied, a bounding robot maintains its stable equilibrium state at a yaw angle of 0○ (perpendicular to the logs), whereas a trotting robot exhibits transitions to yaw angle equilibria of ±θ○ (0○ < θ < 90○). We show that θ can be predicted using the characteristic length of the robot gait and the environment spatial period. |
Monday, March 5, 2018 12:15PM - 12:27PM |
B50.00004: Measuring the Mechanical Coupling in Dynamic Locomotion to Inform Control Izaak Neveln, Simon Sponberg Dynamic locomotion is the result of coordination across distributed subsystems. For example, terrestrial animals maintain coordinated leg patterns to move. To maintain coordination, information is shared through coupling. As coupling magnitude increases, more information is shared globally throughout the subsystems compared to information shared locally, both of which can be measured. In a terrestrial robot, some coupling will be due to mechanical interactions between the legs mediated by the environment and body. As it may be difficult to model these physics, empirical measurements may be necessary to characterize a system to inform control strategies. Using our information measures, we characterize the mechanical coupling of a bounding quadrupedal robot as we alter its non-dimensional inertia, a known parameter of its body-mediated coupling. Global information minus local information is minimized for a particular moment of inertia where impulses from one leg pair have little effect on the other leg pair. Our empirical measurements match simplified analytic and computational models. Because of its applicability to many controlled dynamic systems, this information metric provides the start of a design framework for locomotor control of animals and robots. |
Monday, March 5, 2018 12:27PM - 12:39PM |
B50.00005: Design and assembly of vasoconstriction-like flow regulators for micro channels applications Long Zhou, Yang Zhang The kinematic control of pneumatically-powered robots is still a challenge, partially because of the lack of controllable flow regulators which can be used to adjust the inflation speed. Inspired by the circulatory system, which is controlled not only by the pressure generated by the heart, but also by the vascular smooth muscles that regulate the cross-sectional area of the vessels, we invented a structurally-simple and low-cost flow regulator prototype. Specifically, we introduced a non-cylindrical flexible section in the flow channel. By applying external pressure (squeezing), the flow channel can deform and thus the flow rate can be adjusted. We studied the relations between the flow rate and interference pressure at different vessel pressure and geometries, from which we concluded the most energy efficient combinations of power source and local regulators. This model flow regulator can help improve the accuracy of fluid or air-powered robots by adding flow rates into the control close-loop. We will demonstrate a soft robotic arm we built benefiting from this principle and prototype. |
Monday, March 5, 2018 12:39PM - 12:51PM |
B50.00006: Snake robot’s poor 3-D obstacle traversal reveals snake’s better stability mechanisms Qiyuan Fu, Thomas Mitchel, Nansong Yi, Sean Gart, Chen Li Snakes must often move through complex 3-D terrains like rocks and felled trees with large height changes and wide gaps. Despite snake studies on flat ground, granular media, and branches, the physics of snake-like locomotion in complex 3-D terrain is less known. Recently we discovered that snakes can traverse large steps up to 1/3 body length tall and large gaps up to 1/2 body length wide. Here, we developed a snake robot as a physical model to better understand complex 3-D terrain traversal. The snake robot’s ability to deform its body both laterally and vertically enabled it to achieve similar overall body deformations to that of snakes. A one-direction ratchet wheel mechanism enabled anisotropic frictional profile similar to that of snake scales moving on 2-D surfaces. Despite using these animal kinematics and 2-D friction anisotropy, the snake robot still failed to traverse. As it progressed forward to bridge onto large steps or across large gaps, the robot frequently flipped over (over 90% probability). Closer comparison between the animal and robot revealed that snake’s bottom-wide body cross section shape and ability to locally deform each “segment” for better ground contact via active control and body compliance contributed to its better stability over the robot. |
Monday, March 5, 2018 12:51PM - 1:03PM |
B50.00007: Geometric Swimming with an Extensible Backbone Jacquelin Remaley, Ross Hatton In this work, we consider the effect of curvature-coupled backbone extensibility on the efficiency of swimming systems. We consider two backbones with extensible curvature. One, in which the deformation is piecewise-constant along the body, is inspired by soft pneumatic actuators, while the other encodes a serpenoid motion in which the curvature is a sinusoidal wave traveling along the body. Maximum-efficiency gaits are identified for various levels of extensibility, and the results are compared to non-extensible systems at RMS-equivalent body-lengths. |
Monday, March 5, 2018 1:03PM - 1:15PM |
B50.00008: Collision-Induced Scattering of a Self-Propelled Slithering Robot Jennifer Rieser, Perrin Schiebel, Arman Pazouki, Feifei Qian, Zachary Goddard, Andrew Zangwill, Dan Negrut, Daniel Goldman Our previous work has revealed that the interaction of a snake-like robot with a single row of evenly-spaced vertical pegs rotated the robot trajectory, and that a statistical characterization of these interactions produced a distribution of angles reminiscent far-field wave diffraction: as the spacing between the pegs was decreased, the distribution of scattering angles became multi-modal, with more large angles for smaller spacing. Here, we show that head collisions dominate this rotation, and that, because of the internal driving, they can last for significant fractions of an undulation cycle. Throughout this interaction, the robot head is pushed into but only allowed to slide along the peg surface, causing a rotation of the trajectory by an amount proportional to the interaction duration. Single-peg simulations reveal that this duration is set by the angular position on the peg and the snake undulation phase at impact, with longer-duration collisions occurring near the bottom of the peg. Multi-peg simulations show that large rotations are more frequent when spacing is small, and that the apparent ``interference” be explained by short collisions near the top of a peg becoming inaccessible and preferentially remapped to long collisions near the bottom of an adjacent peg. |
Monday, March 5, 2018 1:15PM - 1:27PM |
B50.00009: Repeated jumping of a single-legged robot on soft ground Paul Umbanhowar, Blake Strebel, Daniel Lynch, Kevin Lynch Achieving high-performance locomotion on yielding substrates such as sand and snow presents substantial challenges for robotic design and control, and to our understanding of the physics of soft ground. To advance these areas, we study the locomotion of a single-leg hopping robot on model ground. The robot consists of two masses, the body and the leg, between which an internal force is applied; the ground consists of particles contained in a fluidized bed, which allows control of the packing density and, consequently, the ground intrusion properties. Unlike rigid or purely compliant substrates, soft ground exhibits hysteresis in the form of tracks created during the stance phase. Consequently, we consider three different cases: 1) jumping in place; 2) gaits in which the leg always encounters undisturbed ground; and 3) gaits where tracks partially overlap. For 1) and 3), we characterize the temporal development of the ground reaction force for repeated leg intrusions, and show that steady states are reached after around five intrusions. Using this knowledge to inform a feed-forward control model, we compare the performance of inter-jump and intra-jump control schemes to achieve fixed height jumps for the three track overlap cases. |
Monday, March 5, 2018 1:27PM - 1:39PM |
B50.00010: Legged Robots Change Locomotor Modes To Traverse 3-D Obstacles With Varied Stiffness Zhiyi Ren, Ratan Sadanand Othayoth Mullankandy, Chen Li When traversing complex 3-D terrains such as grass-like beams, cockroaches often transition between multiple locomotor modes, such as push through, climb over, and roll to maneuver through slits. Here, we use robophysical experiments to study the use of diverse locomotor modes when terrain properties vary and better understand locomotor transitions in nature. We developed a cockroach-inspired legged robot and challenged it to traverse grass-like beams whose torsional stiffness is precisely controlled and systematically varied. We discovered that the robot ran over compliant beam obstacles, quickly rolled to maneuver through medium stiffness beams, but pitch up substantially before rolling to traverse stiff beams. Traversal time increased from 1.0 ± 0.0 s to 3.1 ± 1.2 s as beams became stiffer (P < 0.001, ANOVA). We developed a locomotion energy landscape model to explain these observations, and found that locomotor mode is governed by competition between potential energy barriers from body-terrain interaction and kinetic energy fluctuations from intermittent body/leg-terrain contact. Our study is a major step towards the vision of locomotion energy landscapes as a physics framework to understand and predict locomotor transitions in complex 3-D terrains. |
Monday, March 5, 2018 1:39PM - 1:51PM |
B50.00011: Modeling Multilegged Locomotion: the Friction Dominated Case Dan Zhao, Taesung Heo, Shai Revzen Engineers and biologists often assume non-slip contact for modeling multi-legged locomotion. In our hexapedal robots and cockroaches (Blaberus discoidalis) experiments, a significant fraction of total foot motion is while slipping. Multi-legged contact with slipping poses a modeling challenge: uncertainty about inter-leg force distribution and consequent foot slipping outcomes. Low Reynolds numbers swimmers are governed by a “connection” model. We present a connection model for friction dominated multi-legged locomotion. It takes egocentric leg motions and predicts foot slippage and body center of mass motions. Applied to cockroach data, our model correctly predicts steering directions, but the predicted displacement and curvature of turns are too small. Applied to data from hexapedal robots, our model correctly predicts steering directions under a variety of steering gaits, and is more accurate than that for cockroaches by several metrics. We hope such a model may explain how multi-legged animals control and plan their motions. We also plan to use this reduced order physics models for robot control and design. |
Monday, March 5, 2018 1:51PM - 2:03PM |
B50.00012: The importance of body-limb coordination in a walking tetrapod Baxi Chong, Yasemin Ozkan aydin, Jennifer Rieser, Yunjin Wu, Haosen Xing, Howie Choset, Daniel Goldman Sprawled-posture legged-locomotors (like salamanders) generate movement via cyclic self-deformation using appropriate coordination of limb and body motions. We systematically study how such limb-use patterns (gaits) and body undulation affect locomotor performance in a robophysical and a mathematical model. Our servo-driven salamander-like robot (450 g, 40 cm long) moves on a level bed of poppy seeds using its four limbs, each with two degrees of freedom (up/down and fore/aft), and a joint in the middle of the body which controls horizontal bending. Our mathematical model, which has the same morphology as the robot, extends geometric mechanics [e.g., Hatton, 2013] to legged systems and use granular resistive force theory [Zhang & Goldman, 2014] to model the interaction of the limbs with the ground. The robot and model move using symmetric gaits (i.e., gaits with laterally alternating limb movement) [Hildebrand 1965]. Stride lengths of the robot and the model are in good agreement (± 0.05 body lengths/cycle) over a range of symmetric gaits. Different footfall patterns require different body bending coordination to maximize stride length; the optimal coordination generates stride lengths that are twice as large as that of the worst coordination. |
Monday, March 5, 2018 2:03PM - 2:15PM |
B50.00013: Geometric Gait Optimization Suresh Ramasamy, Ross Hatton Locomotion for many robots and animals is achieved through cyclic changes in shape called gaits. Gait optimization algorithms for locomoting systems must contend with many nonlinear effects, including shape-dependent system dynamics and history-dependent input-output mappings. In this work, we present a variational algorithm for generating optimal gaits for drag-dominated kinematic locomoting systems that encodes recent insights from the geometric mechanics community. For kinematic systems with two shape variables this process is 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 in the case of our algorithm is provided by the flux of the curvature passing through the surface, and surface tension by the cost associated with the gait. We then extend this optimizer to work on systems with 3 and then more than 3 shape variables. We demonstrate this optimizer 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. |
Follow Us |
Engage
Become an APS Member |
My APS
Renew Membership |
Information for |
About APSThe American Physical Society (APS) is a non-profit membership organization working to advance the knowledge of physics. |
© 2024 American Physical Society
| All rights reserved | Terms of Use
| Contact Us
Headquarters
1 Physics Ellipse, College Park, MD 20740-3844
(301) 209-3200
Editorial Office
100 Motor Pkwy, Suite 110, Hauppauge, NY 11788
(631) 591-4000
Office of Public Affairs
529 14th St NW, Suite 1050, Washington, D.C. 20045-2001
(202) 662-8700