Bulletin of the American Physical Society
APS March Meeting 2022
Volume 67, Number 3
Monday–Friday, March 14–18, 2022; Chicago
Session W03: Robophysics IIIFocus Recordings Available
|
Hide Abstracts |
Sponsoring Units: DBIO GSNP Chair: Chen Li, Johns Hopkins University Room: McCormick Place W-176A |
Thursday, March 17, 2022 3:00PM - 3:36PM |
W03.00001: Smart Microscopic Robots Invited Speaker: Paul L McEuen Microscopic robots – robots a few hundred micrometers in size or smaller – have huge potential for applications in numerous fields, from medicine to environmental remediation to studying emergent behavior in robot swarms. However, no previously reported microscopic system can be programmed to accomplish complex tasks autonomously, the dictionary definition of a robot. Here, we report the creation of microscopic robots that meet this standard. We build these robots by integrating electronically controlled microactuators with complementary metal oxide semiconductor (CMOS) electronics. The resulting robots are 100 – 250 micrometers in lateral dimension, are powered by light, and can actuate their legs independently and walk autonomously at speeds greater than 10 micrometers per second without any input of information. We also demonstrate microscopic robots that change their behavior in response to an optical command. This work paves the way for smart microscopic robots that can sense and respond to their environment, receive commands, perform complex functions, and communicate with the outside world. |
Thursday, March 17, 2022 3:36PM - 3:48PM |
W03.00002: Why does a viscous friction ansatz give accurate multi-contact coulomb friction predictions Ziyou Wu, Shai Revzen The complexity of multi-contact mechanics models limits the applications of multi-legged robots. We previously presented a geometric connection model for friction-dominated multi-legged locomotion with slipping using a Coulomb-viscous friction ansatz. Our ansatz has a fast-to-compute closed-form solution, unlike the traditional nonlinear and non-smooth Coulomb model, yet gave accurate predictions for body velocity. To check if the ansatz model predicts the correct friction forces, we collected the robot motion tracking data and ground contact forces from the feet of a hexapod. We found the error lies inside a sector bound, once a single friction coefficient is appropriately chosen. We present some mathematical motivation for why this might occur. |
Thursday, March 17, 2022 3:48PM - 4:00PM |
W03.00003: Biologically inspired actuation via electromagnetic motors Jake E McGrath Electromagnetic motors convert stored energy to mechanical work linearly as described by the force-velocity relationship. In biology, however, muscle actuation supplies power through hyperbolic F-V mechanisms – in which a parameter α characterizes the degree of nonlinearity. Resulting from evolution, nonlinear F-V relationships appear and an optimized α value emerges for specialized muscle tasks. We explore the benefits of implementing biologically inspired actuation in electromagnetic motors by means of a proportional-integral-differential (PID) controller. A PID controller converts the characteristically linear force application of an electric motor to mimic nonlinear mechanical outputs of biological systems. A converted nonlinear electric motor lifts weights of 50-200g, and we record the velocity of the weight and the applied force of the motor. As a proof of concept, we optimize the gain coefficients of the controlling algorithm for a range of input α parameters and, relative to the Hill-type muscle, the nonlinear motor achieves goodness of fit measures of R2 > 0.99. Studies have shown that designing biologically inspired actuators produce comparatively energy efficient systems. We explore the advantages of characterizing a simple linear electric motor with biologically inspired actuation as described by the Hill muscle’s nonlinear F-V relationship. We investigate if there exists an optimized nonlinearity parameter α that provides maximum economic energy consumption. An optimized bio-inspired nonlinear electromagnetic motor manifests robust and energy-efficient mechanical processes. |
Thursday, March 17, 2022 4:00PM - 4:12PM |
W03.00004: Mapping an outdoor odor plume using a mobile chemical sensor Arunava Nag Insects are remarkably adept at tracking wind-borne odor plumes, and understanding their algorithms is of great interest to roboticists. At short distances, walking flies make decisions based on odor statistics such as whiff lengths and encounter frequencies. Experiments in wind tunnels show that these statistics are correlated with distance from the source, and may therefore offer information to flying insects. To test the feasibility of this hypothesis, we measured odor plume statistics driven by natural wind on an open landscape in the Black Rock Desert, Nevada, by traversing a space of 90m2 for 5 hours using a mobile chemical sensor. Stationary wind sensors were arranged around the source to measure the ambient wind speed and direction. Our analysis shows that for higher wind speeds, whiff durations decrease and encounter frequency increases at greater distances from the source. But these trends disappear for lower wind speeds when the wind direction was more variable, suggesting that in real world scenarios plume encounter statistics offer limited information. The statistical relationships we found can be used to develop data-driven plume simulators with realistic dynamics, as well as inform plume tracking algorithms for flying agents such as drones. |
Thursday, March 17, 2022 4:12PM - 4:24PM |
W03.00005: Bulk Electrochemical Actuators for Microscopic Robots and Microscale Medical Tools Jacob Pelster, Qingkun Liu, Wei Wang, Michael F Reynolds, Itai Cohen, Paul L McEuen Actuator energy density defines their applications: a strong microscale actuator can drive microscopic robots with heavy, complex circuitry and would allow for microsurgical tools with an ability to cut through stiff human tissue. We demonstrate microscopic electrically controllable actuators fabricated using CMOS compatible methods exhibiting a larger work density than any offered actuator technology. The devices consist of a bimorph: an expanding Pd layer and passive Ti layer that resists the Pd expansion, generating a moment that results in bending. In an aqueous electrolyte, Pd expands by 3.1% via application of a voltage (~ -1V), catalyzing generation of H that diffuses into the bulk Pd and contracts at ~ 1V, driving out absorbed hydrogen. This work expands on prior actuators utilizing a surface stress, which had limited output energy. The bulk actuators achieve a very small bending radius of 3.8 um when 230 nm thick. Results relating to the measured output force and movement through a gel exhibiting electrochemical and mechanical human tissue properties will be presented. |
Thursday, March 17, 2022 4:24PM - 4:36PM |
W03.00006: Uncovering the mechanisms of wing damage compensation in insect flight using control theory and robophysics Wael Salem, Benjamin Cellini, Heiko D Kabutz, Hari Krishna Hari Prasad, Bo Cheng, Kaushik Jayaram, Jean-Michel Mongeau Flies maintain flight even after losing half of a single wing, a feat with no current analogue in robotics. Compensation is achieved through changes in the motion of the intact and damaged wing. However, the impact of wing damage on performance, and the neuromechanical strategies used to compensate for wing damage, are not well understood. By studying intact and injured flies inside a virtual reality arena, we quantified the impact of wing damage on gaze stabilization. Using system identification techniques, we found that wing damage subtly reduced flight performance. By combining flight data with a robophysical model, we discovered that damage compensation is driven by both active and passive mechanics. The robophysical model and flies exhibited a passive increase in wing amplitude and flapping frequency with increasing wing damage. However, flies decreased the amplitude of the intact wing and shifted the abdomen towards the intact wing, suggesting active control. Using control theory, we show that compensation to wing damage is achieved by active changes in internal gains that trade off stability and performance. Studying the response of flying insects to wing injury can inform the design of flapping-wing robots that maintain function following damage. |
Thursday, March 17, 2022 4:36PM - 4:48PM |
W03.00007: Simultaneous Leg Impacts Lead to a Differentiable Flow Marion Anderson, Shai Revzen Multi-contact dynamical systems are generally thought to have non-smooth properties. Surprisingly, recent work [arXiv:2102.10702] found that for a class of common mechanical systems, the flow will be continuously differentiable even at a multiple collision event with arbitrarily many contacts. We present experimental validation of this finding on a 3-legged hopping robot with springy legs by showing that all 6=3! hopping impact maps are indistinguishable from one another. Using motion capture data of foot position and body pose, we estimate derivative maps for each ordering of foot impacts with the ground. We then compare their pairwise distances on the Grassman Manifold. Finally, we compare prediction errors between the estimated maps. The results suggest that multi-contact locomotion may be easier to model than previously assumed. |
Thursday, March 17, 2022 4:48PM - 5:00PM |
W03.00008: Performance trade-offs in a latch-mediated spring actuated robotic jumper Tanvi Krishnan, Sathvik Divi, Ryan St. Pierre, Sarah Bergbreiter, Mark Ilton Spring-driven mechanisms in robotic systems can be used to drive rapid movements that circumvent the power limitations of small-scale motors, but the integration of motors, latches, and springs into a small jumping system presents a challenge in robotic design. In this work, we experimentally investigate robotic jumpers designed with an integrated latch-mediated spring actuation (LaMSA) system. We explored how trade-offs in system parameters determined the range of optimal performance of the robot. The probability of successful jumps and the measured take-off velocity of the robot depend on the latch radius, spring constant, the gear ratio of the unlatching motor, and the voltage supplied to the motor. We discovered a trade-off between the amount of energy stored in the system and the ability of the system to unlatch, as well as a trade-off between the tunability of the robotic jumper performance and the robustness of the unlatching motor. Our results may allow for future refinement of the robotic jumper and provide us with the ability to tune the whole system performance and control to a desired goal. |
Thursday, March 17, 2022 5:00PM - 5:12PM |
W03.00009: Incorporating a time history of ambient wind improves odor plume tracking success in simulated flying agents TAMZEED ELAHI, Floris van Breugel Insects exhibit exceptionally robust algorithms for navigating airborne turbulent odor plumes while searching for food or mates. There is growing interest in replicating their success with artificial agents. Prior experiments with flying insects in laminar wind tunnels suggest they rely on a “surge and cast” behavior, where the animals turn upwind after encountering the plume, and zigzag in the crosswind direction after losing track of the odor. Despite myriad engineering efforts, the success rate of biological agents remains unmatched, especially under conditions with rapid changes in wind direction. Here, we suggest that an artificial agent can significantly improve its source estimation success rate if it incorporates a memory of wind direction. In our study, we gave our agents a memory of wind direction which it uses along with “surge and cast” to generate a collection of finite candidate routes. With subsequent odor encounters and optimizations, the agent estimates which of these paths is most likely to lead to the actual source. We compared our agent’s performance with the “surge and cast” algorithm and our results suggest that incorporating a time history of wind can substantially improve plume tracking in artificial agents and may also be employed by flying insects. |
Thursday, March 17, 2022 5:12PM - 5:24PM |
W03.00010: Enabling Power and Control Autonomy for Insect Scale Robo-Physical Models William P McDonnell, Kaushik Jayaram Bioinspired insect sized robots can not only emulate the remarkable locomotory capabilities of their animal counterparts, but can also serve as effective "at scale" robo-physical models for discovering novel biomechanical principles and validating their underlying physical mechanisms. The limited applicability of readily available off-the-shelf technologies -- which enable sensing, control and power autonomy at this scale -- is a major factor limiting progress of insect scale robo-physical models and experiments. In this work, we have devised modular and scalable solutions for our insect scale robo-physical model CLARI (compliant legged articulated robot insect). We demonstrate long term untethered operation through wireless energy harvesting and distributed control using custom high voltage smart piezoactuator driver modules. With these developments, we hope to use CLARI to test hypotheses about terradynamic streamlining and energy landscape optimization for locomotion in cluttered terrain, while contributing technologies that foster accelerated progress in the interdisciplinary fields of physics, biology and engineering through open source designs and documentation. |
Thursday, March 17, 2022 5:24PM - 5:36PM |
W03.00011: Accelerating multi-contact modeling using a GPU Advait Deshpande, Ziyou Wu, Shai Revzen Multi-legged locomotion is classically modelled without accounting for contact slippage. Our group has previously demonstrated that such gaits are prone to sizeable slippage, showing evidence from both multi-legged organisms and robots. We have also shown a numerical ansatz based on viscous friction that rapidly provides an approximate solution. Here we report on advances we made in GPU acceleration of this computation, with the goal of demonstrating brute-force trajectory planning for a hexapedal robot with slipping. We search to find the best combination of leg motor commands to achieve a desired body velocity at each time-step - a 6 dimensional search space. We present the use of asynchronous data streams, device-based functions, local memory access and GPU-native sorting in NVIDIA CUDA using the Python Numba framework. Overall, we hope to expand this approach to enable real-time trajectory planning in multi-legged robots. |
Thursday, March 17, 2022 5:36PM - 5:48PM |
W03.00012: Self-organized robotic locomotion by closing the propriosensory feedback loop Bulcsú Sándor, Michael Nowak, Claudius Gros From a dynamical systems viewpoint, robotic locomotion corresponds to attractor trailing in the phase space spanned by the variables describing the state of the controller, the robot's body, and the environment it is placed into. For top-down control the dynamics of the body is driven by a controller signal, allowing for the differentiation between the master and slave subsystems. However, when propriosensory input is also integrated into the controller scheme, self-organized motion patterns emerge due to the local feedback mechanisms. |
Thursday, March 17, 2022 5:48PM - 6:00PM |
W03.00013: Simple Motions as Templates for Generating Gait Spaces of Robophysical Systems Nelson Rosa, Maximilian Raff, C. David Remy A stated goal of the field of robophysics is to understand the core dynamics of complicated robotic systems through simplified models (Aguilar et al., 2016). In the context of legged locomotion, simple often means low-dimensional models whose gaits are used as reference trajectories for robots in the field. An advantage of simple models is the ease with which a set of gaits can be generated. In our work, we expand what simple can mean in terms of gait generation. Our work attempts to mathematically describe the complete set of motions of a broad class of models using fundamental concepts from the fields of nonlinear dynamics and topology. Our results show that many gaits of interest are continuously connected to each other in a properly defined trajectory space. This transforms the problem of gait generation into finding the simplest gait in the connected set for generating the other gaits; we numerically compute the set using continuation methods. This approach has generated a variety of gaits for a legged hopper using a hopping-in-place motion (Gan et al., 2018) and 2D and 3D biped walkers from a standing still position (Rosa and Lynch, 2021) as the simple gaits. These results motivate the study of a model's gait space in simplifying the challenges of gait generation. |
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