Neural dynamics of robust legged robots.

Legged robot control has improved in recent years with the rise of deep reinforcement learning, however, much of the underlying neural mechanisms remain difficult to interpret. Our aim is to leverage bio-inspired methods from computational neuroscience to better understand the neural activity of robust robot locomotion controllers. Similar to past work, we observe that terrain-based curriculum learning improves agent stability. We study the biomechanical responses and neural activity within our neural network controller by simultaneously pairing physical disturbances with targeted neural ablations. We identify an agile hip reflex that enables the robot to regain its balance and recover from lateral perturbations. Model gradients are employed to quantify the relative degree that various sensory feedback channels drive this reflexive behavior. We also find recurrent dynamics are implicated in robust behavior, and utilize sampling-based ablation methods to identify these key neurons. Our framework combines model-based and sampling-based methods for drawing causal relationships between neural network activity and robust embodied robot behavior.

Characterization of organic fouling on thermal bubble-driven micro-pumps.

Thermal bubble-driven micro-pumps are an upcoming micro-actuator technology that can be directly integrated into micro/mesofluidic channels, have no moving parts, and leverage existing mass production fabrication approaches. These micro-pumps consist of a high-power micro-resistor that boils fluid in microseconds to create a high-pressure vapor bubble which performs mechanical work. As such, these micro-pumps hold great promise for micro/mesofluidic systems such as lab-on-a-chip technologies. However, to date, no current work has studied the interaction of these micro-pumps with biofluids such as blood and protein-rich fluids. In this study, the effects of organic fouling due to egg albumin and bovine whole blood are characterized using stroboscopic high-speed imaging and a custom deep learning neural network based on transfer learning of RESNET-18. It was found that the growth of a fouling film inhibited vapor bubble formation. A new metric to quantify the extent of fouling was proposed using the decrease in vapor bubble area as a function of the number of micro-pump firing events. Fouling due to egg albumin and bovine whole blood was found to significantly degrade pump performance as well as the lifetime of thermal bubble-driven micro-pumps to less than 104 firings, which may necessitate the use of protective thin film coatings to prevent the buildup of a fouling layer.

Why animals can outrun robots.

Animals are much better at running than robots. The difference in performance arises in the important dimensions of agility, range, and robustness. To understand the underlying causes for this performance gap, we compare natural and artificial technologies in the five subsystems critical for running: power, frame, actuation, sensing, and control. With few exceptions, engineering technologies meet or exceed the performance of their biological counterparts. We conclude that biology's advantage over engineering arises from better integration of subsystems, and we identify four fundamental obstacles that roboticists must overcome. Toward this goal, we highlight promising research directions that have outsized potential to help future running robots achieve animal-level performance.

Mechanosensory Control of Locomotion in Animals and Robots: Moving Forward.

While animals swim, crawl, walk, and fly with apparent ease, building robots capable of robust locomotion remains a significant challenge. In this review, we draw attention to mechanosensation-the sensing of mechanical forces generated within and outside the body-as a key sense that enables robust locomotion in animals. We discuss differences between mechanosensation in animals and current robots with respect to (1) the encoding properties and distribution of mechanosensors and (2) the integration and regulation of mechanosensory feedback. We argue that robotics would benefit greatly from a detailed understanding of these aspects in animals. To that end, we highlight promising experimental and engineering approaches to study mechanosensation, emphasizing the mutual benefits for biologists and engineers that emerge from moving forward together.

Flies trade off stability and performance via adaptive compensation to wing damage.

Physical injury often impairs mobility, which can have dire consequences for survival in animals. Revealing mechanisms of robust biological intelligence to prevent system failure can provide critical insights into how complex brains generate adaptive movement and inspiration to design fault-tolerant robots. For flying animals, physical injury to a wing can have severe consequences, as flight is inherently unstable. Using a virtual reality flight arena, we studied how flying fruit flies compensate for damage to one wing. By combining experimental and mathematical methods, we show that flies compensate for wing damage by corrective wing movement modulated by closed-loop sensing and robust mechanics. Injured flies actively increase damping and, in doing so, modestly decrease flight performance but fly as stably as uninjured flies. Quantifying responses to injury can uncover the flexibility and robustness of biological systems while informing the development of bio-inspired fault-tolerant strategies.

Strength-mass scaling law governs mass distribution inside honey bee swarms.

To survive during colony reproduction, bees create dense clusters of thousands of suspended individuals. How does this swarm, which is orders of magnitude larger than the size of an individual, maintain mechanical stability? We hypothesize that the internal structure in the bulk of the swarm, about which there is little prior information, plays a key role in mechanical stability. Here, we provide the first-ever 3D reconstructions of the positions of the bees in the bulk of the swarm using x-ray computed tomography. We find that the mass of bees in a layer decreases with distance from the attachment surface. By quantifying the distribution of bees within swarms varying in size (made up of 4000-10,000 bees), we find that the same power law governs the smallest and largest swarms, with the weight supported by each layer scaling with the mass of each layer to the [Formula: see text] power. This arrangement ensures that each layer exerts the same fraction of its total strength, and on average a bee supports a lower weight than its maximum grip strength. This illustrates the extension of the scaling law relating weight to strength of single organisms to the weight distribution within a superorganism made up of thousands of individuals.

Rapid Fabrication of Low-Cost Thermal Bubble-Driven Micro-Pumps.

Thermal bubble-driven micro-pumps are an upcoming actuation technology that can be directly integrated into micro/mesofluidic channels to displace fluid without any moving parts. These pumps consist of high power micro-resistors, which we term thermal micro-pump (TMP) resistors, that locally boil fluid at the resistor surface in microseconds creating a vapor bubble to perform mechanical work. Conventional fabrication approaches of thermal bubble-driven micro-pumps and associated microfluidics have utilized semiconductor micro-fabrication techniques requiring expensive tooling with long turn around times on the order of weeks to months. In this study, we present a low-cost approach to rapidly fabricate and test thermal bubble-driven micro-pumps with associated microfluidics utilizing commercial substrates (indium tin oxide, ITO, and fluorine doped tin oxide, FTO, coated glass) and tooling (laser cutter). The presented fabrication approach greatly reduces the turn around time from weeks/months for conventional micro-fabrication to a matter of hours/days allowing acceleration of thermal bubble-driven micro-pump research and development (R&D) learning cycles.

Spider-Inspired Electrohydraulic Actuators for Fast, Soft-Actuated Joints.

The impressive locomotion and manipulation capabilities of spiders have led to a host of bioinspired robotic designs aiming to reproduce their functionalities; however, current actuation mechanisms are deficient in either speed, force output, displacement, or efficiency. Here-using inspiration from the hydraulic mechanism used in spider legs-soft-actuated joints are developed that use electrostatic forces to locally pressurize a hydraulic fluid, and cause flexion of a segmented structure. The result is a lightweight, low-profile articulating mechanism capable of fast operation, high forces, and large displacement; these devices are termed spider-inspired electrohydraulic soft-actuated (SES) joints. SES joints with rotation angles up to 70°, blocked torques up to 70 mN m, and specific torques up to 21 N m kg-1 are demonstrated. SES joints demonstrate high speed operation, with measured roll-off frequencies up to 24 Hz and specific power as high as 230 W kg-1 -similar to human muscle. The versatility of these devices is illustrated by combining SES joints to create a bidirectional joint, an artificial limb with independently addressable joints, and a compliant gripper. The lightweight, low-profile design, and high performance of these devices, makes them well-suited toward the development of articulating robotic systems that can rapidly maneuver.

Effective locomotion at multiple stride frequencies using proprioceptive feedback on a legged microrobot.

Limitations in actuation, sensing, and computation have forced small legged robots to rely on carefully tuned, mechanically mediated leg trajectories for effective locomotion. Recent advances in manufacturing, however, have enabled in such robots the ability for operation at multiple stride frequencies using multi-degree-of-freedom leg trajectories. Proprioceptive sensing and control is key to extending the capabilities of these robots to a broad range of operating conditions. In this work, we use concomitant sensing for piezoelectric actuation with a computationally efficient framework for estimation and control of leg trajectories on a quadrupedal microrobot. We demonstrate accurate position estimation (<16[Formula: see text] root-mean-square error) and control (<16[Formula: see text] root-mean-square tracking error) during locomotion across a wide range of stride frequencies (10 Hz-50 Hz). This capability enables the exploration of two bioinspired parametric leg trajectories designed to reduce leg slip and increase locomotion performance (e.g. speed, cost-of-transport (COT), etc). Using this approach, we demonstrate high performance locomotion at stride frequencies (10 Hz-30 Hz) where the robot's natural dynamics result in poor open-loop locomotion. Furthermore, we validate the biological hypotheses that inspired the trajectories and identify regions of highly dynamic locomotion, low COT (3.33), and minimal leg slippage (<10%).

Transition by head-on collision: mechanically mediated manoeuvres in cockroaches and small robots.

Exceptional performance is often considered to be elegant and free of 'errors' or missteps. During the most extreme escape behaviours, neural control can approach or exceed its operating limits in response time and bandwidth. Here we show that small, rapid running cockroaches with robust exoskeletons select head-on collisions with obstacles to maintain the fastest escape speeds possible to transition up a vertical wall. Instead of avoidance, animals use their passive body shape and compliance to negotiate challenging environments. Cockroaches running at over 1 m or 50 body lengths per second transition from the floor to a vertical wall within 75 ms by using their head like an automobile bumper, mechanically mediating the manoeuvre. Inspired by the animal's behaviour, we demonstrate a passive, high-speed, mechanically mediated vertical transitions with a small, palm-sized legged robot. By creating a collision model for animal and human materials, we suggest a size dependence favouring mechanical mediation below 1 kg that we term the 'Haldane limit'. Relying on the mechanical control offered by soft exoskeletons represents a paradigm shift for understanding the control of small animals and the next generation of running, climbing and flying robots where the use of the body can off-load the demand for rapid sensing and actuation.

Inverted and vertical climbing of a quadrupedal microrobot using electroadhesion.

The ability to climb greatly increases the reachable workspace of terrestrial robots, improving their utility for inspection and exploration tasks. This is particularly desirable for small (millimeter-scale) legged robots operating in confined environments. This paper presents a 1.48-gram and 4.5-centimeter-long tethered quadrupedal microrobot, the Harvard Ambulatory MicroRobot with Electroadhesion (HAMR-E). The design of HAMR-E enables precise leg motions and voltage-controlled electroadhesion for repeatable and reliable climbing of inverted and vertical surfaces. The innovations that enable this behavior are an integrated leg structure with electroadhesive pads and passive alignment ankles and a parametric tripedal crawling gait. At a relatively low adhesion voltage of 250 volts, HAMR-E achieves speeds up to 1.2 (4.6) millimeters per second and can ambulate for a maximum of 215 (162) steps during vertical (inverted) locomotion. Furthermore, HAMR-E still retains the ability for high-speed locomotion at 140 millimeters per second on horizontal surfaces. As a demonstration of its potential for industrial applications, such as in situ inspection of high-value assets, we show that HAMR-E is capable of achieving open-loop, inverted locomotion inside a curved portion of a commercial jet engine.

Gait studies for a quadrupedal microrobot reveal contrasting running templates in two frequency regimes.

Performance metrics such as speed, cost of transport, and stability are the driving factors behind gait selection in legged locomotion. To help understand the effect of gait on the performance and dynamics of small-scale ambulation, we explore four quadrupedal gaits over a wide range of stride frequencies on a 1.43 g, biologically-inspired microrobot, the Harvard Ambulatory MicroRobot (HAMR). Despite its small size, HAMR can precisely control leg frequency, phasing, and trajectory, making it an exceptional platform for gait studies at scales relevant to insect locomotion. The natural frequencies of the body dynamics are used to identify frequency regimes where the choice of gait has varying influence on speed and cost of transport (CoT). To further quantify these effects, two new metrics, ineffective stance and stride correlation, are leveraged to capture effects of foot slippage and observed footfall patterns on locomotion performance. At stride frequencies near body resonant modes, gait is found to drastically alter speed and CoT. When running well above these stride frequencies we find a gait-agnostic shift towards energy characteristics that support 'kinematic running', which is defined as a gait with a Froude number greater than one with energy profiles more similar to walking than running. This kinematic running is rapid (8.5 body lengths per second), efficient (CoT = 9.4), different from widely observed SLIP templates of running, and has the potential to simplify design and control for insect-scale runners.

Cockroaches traverse crevices, crawl rapidly in confined spaces, and inspire a soft, legged robot.

Jointed exoskeletons permit rapid appendage-driven locomotion but retain the soft-bodied, shape-changing ability to explore confined environments. We challenged cockroaches with horizontal crevices smaller than a quarter of their standing body height. Cockroaches rapidly traversed crevices in 300-800 ms by compressing their body 40-60%. High-speed videography revealed crevice negotiation to be a complex, discontinuous maneuver. After traversing horizontal crevices to enter a vertically confined space, cockroaches crawled at velocities approaching 60 cm⋅s(-1), despite body compression and postural changes. Running velocity, stride length, and stride period only decreased at the smallest crevice height (4 mm), whereas slipping and the probability of zigzag paths increased. To explain confined-space running performance limits, we altered ceiling and ground friction. Increased ceiling friction decreased velocity by decreasing stride length and increasing slipping. Increased ground friction resulted in velocity and stride length attaining a maximum at intermediate friction levels. These data support a model of an unexplored mode of locomotion--"body-friction legged crawling" with body drag, friction-dominated leg thrust, but no media flow as in air, water, or sand. To define the limits of body compression in confined spaces, we conducted dynamic compressive cycle tests on living animals. Exoskeletal strength allowed cockroaches to withstand forces 300 times body weight when traversing the smallest crevices and up to nearly 900 times body weight without injury. Cockroach exoskeletons provided biological inspiration for the manufacture of an origami-style, soft, legged robot that can locomote rapidly in both open and confined spaces.

Mechanical processing via passive dynamic properties of the cockroach antenna can facilitate control during rapid running.

The integration of information from dynamic sensory structures operating on a moving body is a challenge for locomoting animals and engineers seeking to design agile robots. As a tactile sensor is a physical linkage mediating mechanical interactions between body and environment, mechanical tuning of the sensor is critical for effective control. We determined the open-loop dynamics of a tactile sensor, specifically the antenna of the American cockroach, Periplaneta americana, an animal that escapes predators by using its antennae during rapid closed-loop tactilely mediated course control. Geometrical measurements and static bending experiments revealed an exponentially decreasing flexural stiffness (EI) from base to tip. Quasi-static experiments with a physical model support the hypothesis that a proximodistally decreasing EI can simplify control by increasing preview distance and allowing effective mapping to a putative control variable--body-to-wall distance--compared with an antenna with constant EI. We measured the free response at the tip of the antenna following step deflections and determined that the antenna rapidly damps large deflections: over 90% of the perturbation is rejected within the first cycle, corresponding to almost one stride period during high-speed running (~50 ms). An impulse-like perturbation near the tip revealed dynamics that were characteristic of an inelastic collision, keeping the antenna in contact with an object after impact. We contend that proximodistally decreasing stiffness, high damping and inelasticity simplify control during high-speed tactile tasks by increasing preview distance, providing a one-dimensional map between antennal bending and body-to-wall distance, and increasing the reliability of tactile information.