Bulletin of the American Physical Society
APS March Meeting 2023
Volume 68, Number 3
Las Vegas, Nevada (March 5-10)
Virtual (March 20-22); Time Zone: Pacific Time
Session W10: Robophysics IIIFocus
|
Hide Abstracts |
Sponsoring Units: DBIO Chair: Nick Gravish, University of California, San Diego Room: Room 202 |
Thursday, March 9, 2023 3:00PM - 3:36PM |
W10.00001: Fish-inspired robot manipulation Invited Speaker: Hannah S Stuart Fish demonstrate great dexterity as they feed in both the water column (floating prey) and on substrates (fixed prey). We are working across multiple projects to understand and apply fish feeding-inspired mechanisms to enhance the manipulation abilities of underwater robots. In suction feeding, fish capture floating prey by pulling water into their mouth. We employed suction flow at the fingertips of a humanoid hand and found that gentle suction conditions, on the order of 1N of force, improve the rapid capture of floating objects up to 10 kg. To study feeding on stationary prey, we measured the forces and motions of the sunburst butterflyfish (Chaetodon kleinii) as it took bites off a fixed, soft prey. We observed characteristic behaviors of the fish, including a lateral head-flick for prey shearing and tearing. We constructed a robophysical model of the sunburst butterflyfish in order to characterize the effect of this head-flick action. In both robophysical projects, we found that simple template models in dynamic simulation provide accurate predictions of robot behaviors. This surprising simplicity supports rapid parametric study of the effects of suction and body motion in these cases. |
Thursday, March 9, 2023 3:36PM - 3:48PM |
W10.00002: A Shannon-inspired framework for multi-legged matter transport Baxi Chong, Juntao He, Daniel Soto, Tianyu Wang, Daniel Irvine, Daniel I Goldman During the last century, Shannon predicted that signals could be reliably transmitted over noisy channels based on digital redundant coding in the form of binary bit sequences. Such prediction has become increasingly critical in modern communication. Here, we establish an analogous framework to describe matter transport, in the manner of robot locomotion, over noisy landscapes. The key idea is to digitize thrust-generation via basic active contacts, bacs. Theory and experiments demonstrate that appropriate temporal and/or spatial bac generation (via serially-connected modular robots) can facilitate reliable transport of multi-legged systems on laboratory and field rugose terrains without the need for complex sensing/control. We hypothesize that the analogy between signal and matter transportation can accelerate development of devices with mobility approaching that of living systems. |
Thursday, March 9, 2023 3:48PM - 4:00PM |
W10.00003: Emergent gait transitions of a quadrupedal robot under obstacle modulation Xingjue Liao, Feifei Qian Biological gaits are results of both neural control signals and mechanical feedback. Recent biological studies (Wilshin et al., 2017) demonstrated that when moving from flat ground to rough terrains, dogs would shift from walk-like gaits to trot-like gaits. Inspired by this result, we use a quadrupedal robot as a robophysics model, to study how a ''soft’’ feedforward gait is affected by repeated ''phase perturbations’’, as the robot traverses an array of evenly spaced half-cylindrical obstacles. We modelled the feedforward quadrupedal gait as relative phases between the leg pairs, (Φ1, Φ2, Φ3), with a ''gait stiffness'', K. Here Φ1, Φ2, Φ3 denote the phase differences between the left and right front legs, left rear and right front legs, right rear and right front legs, respectively. Upon each leg-obstacle collision, a phase perturbation, L, is applied to the contacting robot leg, causing the leg to delay or advance its phase based on the contact position on the obstacle, and therefore modulates the robot gait. Simulation results with a feedforward ''trot’’ gait, (Φ1, Φ2, Φ3) = (0.5,0,0.5), suggested that with a relatively small phase perturbation as compared to the gait stiffness, the robot could converge to a new stable gait under repeated phase perturbations. Two distinct groups of converged gaits were observed: (i) (0.5-C1, 0, 0.5-C1), and (ii) (0.5,1-C2, 0.5-C2), where C1 and C2 are primarily governed by the ratio of the phase perturbation, L, to the gait stiffness, K. |
Thursday, March 9, 2023 4:00PM - 4:12PM |
W10.00004: RoboPhysics inspired experiments to understand how canines move over regular obstacle arrays Benjamin Doshna, Michelle Joyce, Haodi Hu, Feifei Qian, Simon Wilshin, Andrew J Spence Animals move with economy and dexterity that far exceed their robotic counterparts, although robots appear to be catching up. Combined Robophysical and animal experiments have the potential to reveal general principles of legged locomotion. Recent robot experiments have shown that changes in gait can have large effects on locomotor behavior when moving over regular obstacle arrays. For example, changing gait between trot and bound alters the stable heading of travel over the obstacles. Inspired by this, we hypothesized that canines would strongly adapt their gait when confronted with similar obstacles. Surprisingly, they did not exhibit large changes in gait. Therefore, in this study we step back to characterize changes in basic stride parameters and spatial aspects of locomotion. These include stride length, heading relative to the obstacles, and comparisons to movement over flat ground. We report results from 9 dogs moving over domed obstacles with spacing that brackets their natural trotting stride length (80, 100, and 120 percent). These studies may reveal how legged systems can utilize their interaction with complex environments for steering, course planning, and/or more stable and energetically efficient movement. |
Thursday, March 9, 2023 4:12PM - 4:24PM |
W10.00005: Using an insect-scale robot to explore the impact of limb number and length on locomotion in unstructured terrain. Perrin E Schiebel, Velin H Kojouharov, Robert J Wood Hexapod insects and quadruped lizards ably traverse unstructured terrains. Using quadruped and hexapod versions of the insect-scale Harvard Ambulatory Microrobot as a robophysical model, we endeavored to elucidate how limb morphology and kinematics impact locomotor performance in rugged terrain as a function of the number of limbs. Guided by our previous work studying passive compliant limbs in small-scale locomotors, we varied the amount of the limb in contact with the substrate by incorporating passive joints to partially collapse the limb. We placed pairs of limbs of different lengths on the robots and measured speed and mechanical cost-of-transport as they traversed a 3D printed terrain with randomized feature heights. We tested the hexapod using an alternating tripod gait and the quadruped using both a walk—the quadrupedal analog to the alternating tripod—and a trot. We included limb and gait combinations which result in body-terrain contact to explore whether using the body as an additional point of support could help ensure robust locomotion. |
Thursday, March 9, 2023 4:24PM - 4:36PM |
W10.00006: Terrestrial swimming in multilegged robots Juntao He, Baxi Chong, Shengkai Li, Eva Erickson, Kelimar Diaz, Tianyu Wang, Daniel Soto, Daniel I Goldman Locomotion is typically studied in continuous media where bodies and legs experience forces generated by the flowing medium, or on solid substrates dominated by friction. In the former, slipping through the medium is unavoidable and necessary for propulsion. In the latter, slip is often assumed undesirable and/or minimal. We discover in laboratory experiments that terrestrial locomotion of a multi-legged robophysical model (L = 20 to 160 cm, 2 to 8 segments) proceeds via slipping and can thus be viewed as terrestrial "swimming". Robophysical experiments varying leg and body wave amplitudes reveal that inertial effects are minimal. Numerical simulations and theoretical analysis demonstrate that periodic leg-substrate contact allows for analysis via an effective Resistive Force Theory akin to that in flowable media where the system experiences an effective viscous drag and acquired drag anisotropy. Analysis of body and leg dominated propulsion reveals body undulation can buffer the robot against vertical heterogeneities (randomly distributed leg scale vertical obstacles), which extends the terrestrial swimming capabilities of the robot beyond flat frictional terrain. |
Thursday, March 9, 2023 4:36PM - 4:48PM |
W10.00007: Multi-robot connection towards collective obstacle field traversal Haodi Hu, Elliott Meeks, Feifei Qian Environments filled with large rocks and boulders can be challenging for legged locomotion. Inspired by ants collectively traversing obstacles through physical connections with each other, we study the locomotion performance of a group of "connectable'' robots over an obstacle field filled with half-cylindrical "boulders''. Each robot is extremely simple, with a cuboid body and one rotary motor actuating four vertical peg legs that move in pairs. The robots could connect to one another through the side of their cuboid bodies. Simulation results suggested that individual robots could not advance well within the obstacle field on their own (less than 0.75 body length (BL) over 15 stride cycles). However, by connecting to a neighboring robot, the two-robot system could begin to produce continuous displacement (up to 1.5 BL per stride cycle) across the obstacle field. Using an "obstacle disturbance selection'' framework, we modeled obstacle disturbances exerted on each individual robot leg as a function of obstacle inclination at the contact positions. The model suggested that by connecting to different neighbors, the two-robot system could produce collective translational or rotational movements through the utilization of obstacle disturbances. |
Thursday, March 9, 2023 4:48PM - 5:00PM |
W10.00008: Force sensing can help robots reconstruct potential energy landscape and guide locomotor transitions to traverse large obstacles Yaqing Wang, Chen Li Legged robots already excel at maintaining stability during upright walking and running to step over small obstacles. However, they must further traverse large obstacles comparable to body size to enable a broader range of applications like search and rescue in rubble and sample collection in rocky Martian hills. Our lab's recent research demonstrated that legged robots can traverse large obstacles if they can be destabilized to transition across various locomotor modes. When viewed on a potential energy landscape of the system, which results from physical interaction with obstacles, these locomotor transitions are strenuous barrier-crossing transitions between landscape basins. Because potential energy landscape gradients are conservative forces, we hypothesize that sensing obstacle interaction forces allows landscape reconstruction, which can guide robots to cross barriers at the saddle to make transitions more easily (analogous to crossing a mountain ridge at its saddle). Here, we created a robophysical model with custom 3-axis force sensors and surface contact sensors to measure forces during interaction with large obstacles. We found that the measured forces indeed well captured potential energy landscape gradients. Our next step is to understand how to use the locally measured gradients to sufficiently reconstruct the potential energy landscape so that saddles can be identified. This will enable robots to make locomotor transitions at the landscape saddle with the lowest resistance. |
Thursday, March 9, 2023 5:00PM - 5:12PM |
W10.00009: Connection-Based Data-Driven Gait Modeling of a Quadruped Ziyou Wu, Shai Revzen We have recently shown that connection-based models arising from geometric mechanics apply to legged systems, both biological and robotic, whether they slip or maintain non-slip contacts with the substrate. A key assumption that underlies these physics – that friction annihilates momentum quickly – breaks down for large trotting quadrupeds, raising the question of how well data-driven connections approximate their observed motion. We report initial results from a Ghost Robotics Vision 60 robot, measuring ego motion using iterative closest point (ICP) estimation from LIDAR data to solve for the body velocity, and using encoder data for body shape. In its trotting gait the robot approximately conserves momentum around the line connecting its stance feet, yet the connection term of the reconstruction equation, which ignores momentum, accounted for 97% of observed forward body velocity. The Vision 60 system will allow us to test how well the connection captures the physics of real-world surfaces and motions. |
Thursday, March 9, 2023 5:12PM - 5:24PM |
W10.00010: An Autonomous Tensegrity Robot with Closed-loop Control and Real-time State Estimation William R Johnson Tensegrity robots are lightweight, actuated structures composed of rigid struts and elastic tendons that can accommodate extreme deformations. Their ability to survive harsh impacts through deformations motivates tensegrity robots for navigating unstructured environments. However, tensegrity robots are limited by insufficient onboard sensing and a large sim2real gap. Here, we present an autonomous 3-bar tensegrity robot with elastic sensor tendons for closed-loop control and sim2real transfer. Highly deformable sensors for feedback control allow this robot to achieve high speeds (34 body lengths per minute), climb the steepest incline (28 degrees) of any tensegrity robot, navigate unstructured terrains (pebbles, grass, sand, etc.), and accurately reconstruct its shape and orientation (within 10% of the bar length) relative to Earth’s gravitational and magnetic fields. Finally, we demonstrate high-fidelity transfer of control policies from a differentiable physics-based simulation to the real robot in a real2sim2real pipeline where the physical parameters in the simulation are repeatedly updated based on data from the real robot. |
Thursday, March 9, 2023 5:24PM - 5:36PM |
W10.00011: Compliant cable-driven limbless robot for complex terrain navigation Velin H Kojouharov, Tianyu Wang, Christopher J Pierce, Kelimar Diaz, Baxi Chong, Daniel I Goldman
|
Thursday, March 9, 2023 5:36PM - 5:48PM |
W10.00012: Lattice transport via mechanical intelligence in undulatory locomotors Tianyu Wang, Christopher J Pierce, Velin H Kojouharov, Kelimar Diaz, Baxi Chong, Hang Lu, Daniel I Goldman Limbless locomotion emerges across length scales and within complex environments. Theoretical and robotic models have emphasized body kinematics and neuro/electronic control, but these approaches are far from reproducing the performance of even seemingly simple limbless organisms like tiny nematode worms. Our observations of nematode locomotion in obstacle lattices suggest that many collision-instigated behaviors occur passively, thus we hypothesize that the bilateral muscle actuation mechanism provides favorable passive mechanics for heterogeneity navigation. To test the hypothesis, we implemented the bilateral actuation morphology in a limbless robophysical model (L=86cm, W=7cm), and the limbless robot demonstrated unprecedented open-loop mobility (high speed, low energy cost) in lattices (40cm, 20cm, and random spacing) that approaches nematodes. The limbless robot also displayed analogous passive behaviors as we observed in nematodes. The highly damped nature of the locomotion allows the meter-scale robot model to reveal the importance of mechanical intelligence in C. elegans in complex environments. We further implemented a nematode mechanosensation-inspired closed-loop strategy that reinforces the robot's inherent mechanical intelligence and augments locomotor performance. Our robophysical approach suggests a fruitful modeling approach to understanding the relative roles of mechanical intelligence and active control in limbless animal behavior and neurobiology. |
Thursday, March 9, 2023 5:48PM - 6:00PM |
W10.00013: Robophysical modeling of dynamic posture modulation of vibration sensing in orb weaving spiders Eugene Lin, Yishun Zhou, Luke Moon, Andrew Gordus, Chen Li Orb-weaving spiders are functionally blind and detect prey-generated web vibrations through vibration sensors at their leg joints to locate and identify prey caught in their (near) planar webs. Previous studies focused on how spiders use web geometry, silk properties, and web pre-tension to modulate vibration sensing. Spiders can also dynamically adjust their posture while sensing prey, which may be a form of active sensing (Hung, Corver, Gordus, 2022, APS March Meeting). However, whether this is true and how it works is poorly understood, due to difficulty of measuring the dynamics of the entire prey-web-spider interaction system all at once. Here, we developed a robophysical model of the system to test this hypothesis of active sensing and discover its principles. Our model consists of a vibrating prey robot and a spider robot that can adjust its posture, with torsional springs at leg joints and accelerometers to measure joint vibration. Both robots are attached to a physical web made of cords with qualitatively similar properties to real spider web threads. Load cells measure web pre-tension and a high-speed camera system measure web vibrations and robot movement. Preliminary results showed vibration attenuation through the web from the prey robot. We are currently studying the complex effects of spider robot’s dynamic posture change on vibration propagation across the web and leg joints, by systematically varying the parameters of prey robot vibration, spider robot leg posture, and web pre-tension. |
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