Session B14: Robophysics: Robotics Meets Physics I: Flight & Swimming

Live

Living and robotic systems share the need and the challenge of navigating complex environments. Notable natural examples are provided by the tracking of surface-bound trails of odor cues, long-distance orientation by using airborne olfactory cues, and flight in the lowest layers of the atmosphere. On the biological side, terrestrial animals, insects, and birds evolved navigation strategies that accomplish the above tasks with an efficiency that is often surprising and yet unmatched. On the robotic side, olfactory sniffers and unmanned aerial vehicles face similar challenges. In addition to basic scientific motivations, technological applications range from the automated location of explosives, chemical, and toxic leaks, to the monitoring of biodiversity, surveillance, disaster relief, cargo transport, and agriculture. The interdisciplinary interplay between biology, physics, and robotics is key to jointly advancing fundamental understanding and technology. I shall review the above natural phenomena, then discuss the physics that constrains and shapes the navigation tasks, and conclude with the relevant strategies.   

Live

Centimeter scale biological fliers have adopted two seemingly distinct strategies for high-power, high-frequency flight. Some use time-periodic drive of resonant mechanical systems, termed synchronous flight. Others achieve self-excited, “asynchronous” oscillations via delayed stretch activation (dSA) in their flight muscles. dSA is a delayed rise in force following stretch that allows a pair of opposing dSA muscles to produce emergent limit cycle behavior. Here we unify the periodic drive and dSA activity of flight muscle under one framework. We first identify evolutionary transitions between the two strategies, suggesting that insects can transition from one regime to the other. We then show that synchronous moth muscle unexpectedly also has dSA, but at a scale insufficient to lead to emergent self-excitation. To address how discrete oscillatory behavior emerges from a continuum of actuator properties, we combined the two types of actuation with a resonant spring-wing mechanics model. In simulation, we show that the same dynamical system can produce both synchronous and asynchronous oscillations divided by an intermediate regime with no coherent oscillations. The traditional dichotomy of synchronous and asynchronous insect is therefore two regimes of a spring-wing dynamics.   

Live

Flapping-wing insects are thought to have adopted one of two flight muscle actuation modes: periodic forcing of a resonant wing system, termed synchronous flight, and self-excited, asynchronous, forcing via delayed stretch activation (dSA) in flight muscles. Recently, observations have identified dSA in the flight muscle of hawkmoth (Manduca sexta), a synchronous flyer, suggesting that properties of synchronous and asynchronous modes may exist in the same muscle. We leveraged simulation and two robophysical spring-wing systems - one dynamically-scaled and one at insect scale - to study the interplay between synchronous forcing and asynchronous self-excitation in the same physical system. In this talk, we detail the construction and implementation of a self-excited spring-wing model based on measurements of insect muscle response. We show that wing kinematics suitable for flight only arise in either strongly synchronous or strongly asynchronous regimes. Interference between different oscillation timescales induces period-doubling and other effects that create a “valley of death” for flight performance in the intermediate regime. These results provide dynamical arguments for the sharp differentiation in insect phylogeny between synchronous and asynchronous insect flight actuation.   

Live

Many flapping wing insects use muscle pairs in which a stretch of the muscle causes a delayed-stretch activation (dSA) force without neural input. Delayed-stretch activation between antagonistic muscle pairs thus sets up self-excited, emergent wing oscillations without the need for a neural “clock”. Despite substantial work on the physiology and neuroscience of dSA muscles, a broader systems level understanding of the dynamics of dSA actuation are lacking. I will present our theoretical, numerical and robophysical modeling of dSA spring-wing systems to establish a parameter space of emergent dSA oscillations. We focus on two factors in dSA actuation, the time-delay τ between muscle stretch and peak force, and the amplitude of the force response, F_SA. Physiology experiments have observed a linear relationship between \tau and wingbeat period across a broad range of asynchronous insects, yet the consequences of τ are unclear. Our experiment and analysis demonstrate the importance of delay in dSA actuation with too much delay inhibiting emergent oscillations, and too little delay being impractical from physiological and dynamical limitations. Overall in this talk we will establish the rules of dSA oscillations in a model system of asynchronous spring-wing systems.   

Live

In recent years, there has been heightened interest in developing mm-scale flying vehicles, both for at-scale studies of insect flight and collective behaviors, and for applications in environmental monitoring, structural repair, search & rescue, and archaeological studies. However, at such small scales it is difficult to obtain sufficient system performance to allow for autonomous operation. In this talk we will describe several advances for the Harvard “RoboBee” platform (an 80-mg flapping-wing vehicle) over the past five years that are bringing this goal closer to reality, including non-linear resonance modeling, multi-wing designs, and scaling analysis. These advances contributed to the sustained untethered flight of a 90-mg four-wing vehicle carrying 170 mg of electronics and solar cells. While still lacking onboard control and sensing, this vehicle demonstrated thrust-per-muscle-mass and system thrust-efficiency matching that of typical biological counterparts such as bees, and the modeling framework shows promise for surpassing these metrics in the future.   

Live

Animals, like many robots, are constantly in motion. As they move, their own movements shape the information that is acquired by their sensory systems. Unexpected perturbations, such as a gust of wind knocking a flying insect off course, complicates not just locomotion, but sensing itself. We studied how flying fruit flies overcome these challenges by actively controlling their head to shape the visual inputs that are sensed by the eyes. We revealed that head movements rapid slowed down visual disturbances, thereby reducing motion blur and allowing flies to encode visual motion speeds more than twice as large as current models of insect vision purport. Wing steering efforts followed head movements by 30 ms, revealing a temporal order where the head filters visual information which then shapes downstream wing steering efforts. By comparing the responses of head-free and head-fixed flies, we revealed that head movements increased the strength and power of downstream wing steering efforts and improved coordination between the wings and visual input, demonstrating that active sensing can enhance motor control. Our results provide inspiration for the control of an active sensor on a moving body in robotic visual guidance systems.   

Live

Exploring unknown underwater environments containing high densities of obstacles is a significant challenge for swimming robots. Many swimming animals can adapt to complex underwater terrain and swim through the clutter with their maneuverability and compliant bodies. Upon that premise, we designed a bioinspired two DOF soft swimming robot with high maneuverability and passive adaptability to obstacles. This swimming robot mimics the compliant vertebrate structure of biological swimmers and, with just two actuators, can not only perform basic maneuvers like forwarding locomotion (19.35 cm/s) and yaw/pitch turning (44.09 deg/s), but can also perform a novel roll rotation maneuver (107.78 deg/s). Additionally, the flexible ‘vertebrate’ and foldable fins allow the passive mechanical adaptability required to pass through narrow passages or cluttered terrain. We studied the robot’s obstacle passing capabilities among several underwater obstacles, including narrow upright poles, planar orifice, linear channel, and a set of tunnels. Even without closed-loop control, the robot passed through each obstacle configuration relatively efficiently. Such high maneuverability and passive adaptability would greatly enhance swimming robot locomotion robustness in complex, cluttered environments.   

Live

Studying the swimming strategy of creatures could help to design bio-robots. Here, we display a unique swimming style of larva of Chironomus plumosus. The larva with a length about 14mm uses an interesting ‘figure-of-eight’ swimming style during the escape movement, that is, the body periodically bends into a circle and then fully unfolds for propulsion. We use a 3D model combining with computational fluid dynamics to study the kinematics and dynamics of larval locomotion and found that larva not only moves forward but also rotates its body around the center of mass. Specially, body rotation is important for swimming faster, which influences the net force orientation. Wavelength(λ) of curvature wave, body deformation frequency(ω) and Reynolds number(Re) could influence body rotation angle. It is found that there is an optimal angle range which makes the travel path straighter and makes larva move faster. Therefore we can control larval swimming by adjusting λ, ω and Re to make body rotation angle be in the optimal range. Our study provides a better understanding on the hydrodynamics of the larva and guidance for the design of underwater robots at millimeter scale.   

Live

Microorganisms with appendages (e.g., flagella) rely on various strategies to generate locomotion in highly viscous environments. Despite possessing similar morphology, different quadriflagellate algae species consistently self-propel at different speeds. Here to test if these performance differences are a sole function of the diverse gaits (appendage coordination patterns) employed by different species, we developed a macroscopic robophysical model (four appendages, body length of 3.9 cm) with the capability to self-propel in a viscous fluid (mineral oil, 1,000 cSt). We tested swimming performance in three distinct gaits, the pronk, the trot, and the gallop, and tested the effects of appendage orientation relative to the cell body. With perpendicular appendages, the robot achieved a speed of 0.020-0.1 body lengths per second depending on the gait, with the trot displaying the greatest speed. Robot performance was comparable to the algae across gaits. With parallel appendages, swimming performance decreased significantly for all gaits. Our results show a minimal robophysical modeling can aid our understanding of control principles of low Reynolds swimming in biological systems and inspire design of autonomous microrobots.   

Live

Animals use two main platforms for fluid transport: pressure driven flows through vascular networks and cilial transport. Microfluidics mimics the first strategy. We still don’t have good platforms for mimicking the second strategy. Nevertheless, such a platform is worth pursuing since individually addressable micro-scale robotic cilia could enable unprecedented control over microfluidic environments. For example, they could sort microscale particles, control chemical reactions, and transport viscoelastic materials. Here, we present a new platform for fabricating artificial cilia that uses surface electrochemical actuators made of nm thin Platinum films. Because these actuators are driven asymmetrically via cyclical application of ~0.2 V to the Pt, these robotic cilia can be integrated with control circuits and power sources, allowing for sequential and addressable generation of arbitrary flow patterns. I will present our attempts to make such an artificial cilia chip and integrate it with a data acquisition device to build a microfluidic chip that can generate reconfigurable micro fluidic patterns and is controllable via a computer. We envision that this technology will usher in unparalleled control of complex fluids moving over surfaces.   

On Demand

In this work, we built a Robo-physical model to study the effect of various physical, geometrical, and kinematic parameters of fin on performance. The model can swim along a large circular path continuously to carry extensive experiments. We examined 12 fins in total, which were a combination of 3 different shapes, 3 flexibilities, and uniform or tapered thickness for a particular shape. Each fin was tested with 4 different frequencies and two different amplitudes of oscillation. The results of 96 trials revealed: 1) for any frequency of oscillation there is optimal flexibility that minimizes the cost of transportation. 2) The swimming speed linearly varies with the tail tip velocity. 3) Rectangular fin with the same aspect ratio is more efficient compared to the remaining two shapes 4) Uniformly thick fins require less work and torque while a tapered fin generates more speed. Furthermore, a numerical model was developed using panel method and free vortex sheet. We used this model to find the optimum phase offset between heave and pitch motion in a freely swimming fish and compared that to a tethered fish model. The results were significantly different for tethered and freely swimming cases.