HAMR-Jr Is a Speedy Quadrupedal Robot the Size of a Penny

The last time we checked in with the Harvard Ambulatory MicroRobot (HAMR) was in 2018, when I spent far too much time trying (with a very small amount of what might charitably be called success) to adapt some MC Hammer lyrics for an article intro. Despite having micro robot" right in the name, if we're talking about insect scale, HAMR was a bit chunky, measuring about 5 centimeters long and weighing around 3 grams. At ICRA this week, we've been introduced to a new version of HAMR, called HAMR-Jr, which is significantly smaller: just a tenth of the weight, and comes up to about knee-high on a cockroach.

HAMR-Jr may be tiny, but it's no slouch-piezoelectric actuators can drive it at nearly 14 body lengths (30 cm) per second, at a gait frequency of 200 hertz. The actuators can be cranked up even more, approaching 300 Hz, but the robot actually slows down past 200 Hz, because it turns out that 200 Hz hits a sort of resonant sweet spot that gives the robot as much leg lift and stride length as possible.

It's worth mentioning that even the fastest legged insects don't reach a gait frequency of 200 Hz. While the Australian tiger beetle seems to be the world's fastest legged insect, able to reach about 2.5 m/s (!) when chasing prey with a gait frequency of just a few tens of hertz, what's more relevant here is probably the fastest insect relative to its size (so, fastest speed measured in body lengths per second). That award goes to a tiny species of mite found in California, able to run at nearly 200 body lengths per second. This works out to something like 0.25 m/s because the mites are seriously tiny (sesame-seed sized), but juveniles managed to hit a stride frequency of 135 Hz, which is the same ballpark as HAMR-Jr, albeit on a much smaller scale.

Since I'm totally lost in this insect muscle frequency topic now, the reason that 100-ish Hz seems to be about the limit for insect gait frequencies is that you start running up against biological limits of motor nerve impulses and muscle fiber activation. The absolute highest frequency of synchronized insect muscle contractions is 224 Hz, taking place in the tymbal muscles of the Australian cicada, which is what it uses to make that cicada noise. The word synchronized" implies that you've got nerve impulses and muscles working together, but you can also have nerve impulses and muscles working asynchronously, which takes a lot less active coordination and can bypass some of those biological control limits. Some insects use asynchronous arrangements of muscles acting antagonistically to cause their bodies to flex at extremely high frequencies, usually to power their wings. Biting midges use this technique to beat their wings at 1046 Hz, and even this absurdly high frequency is heavily constrained by the wings themselves, and removing most of the wing area allows what's left to flap at over 2200 Hz.

With HAMR-Jr, we can cycle the legs at the rate of over 200 Hz, which is unprecedented for terrestrial biological systems, and we did not have a priori notion about the locomotion dynamics at these high stride frequencies" -Kaushik Jayaram, University of Colorado Boulder

Anyway, let's get back to HAMR-Jr, shall we? The robot is able to trot, pronk, bound, and jump, and can also move sideways like a crab. With payloads up to its own mass (320 mg), the performance of the robot doesn't change much, which is a good indicator for its ability to handle a payload that could include a battery and some sensors. There's plenty of room for improvement here, though, and some exploratory testing suggests that a more rigid structure could HAMR-Jr's payload to at least 3.5 g.

As for as raw speed is concerned, the biggest performance hit caused by scaling down from the larger version of HAMR is likely the result of decreased mass and inertia, causing a lot of slippage of HAMR-Jr's lil' feet. One way of solving this could be with tiny grippy shoes of some kind, and it also seems like trading off stride frequency for increased stride length could also increase locomotion speeds.

For more details, we spoke via email with lead author Kaushik Jayaram, assistant professor of robotics and systems design, at the University of Colorado, in Boulder, to find out (among other things) whether at some point we'll be seeing HAMR-Mini, HAMR-Nano, and HAMR-Atto. And if you'd like to ask the authors a question yourself, the ICRA info page and Slack discussion channel for this paper can be found here.

IEEE Spectrum: What was the trickiest thing you ran into while scaling HAMR-VI down into HAMR-Jr?

Kaushik Jayaram: The design and fabrication process was the easiest part of this work because we did not have to make any changes to our workflow. The trickiest challenges we encountered were mostly practical. For example, we are fast approaching the tolerance limits of the commercially available sheets of raw materials, and small variations in material (elastic modulus) or geometric properties (thickness) or glue while assembly can have a significant impact on robot performance. Also, handling small and fragile parts and folding them under a microscope required a lot of patience and finger nimbleness.

The other tricky thing was having a guarantee about locomotion performance. For example, with HAMR-Jr, we can cycle the legs at the rate of over 200 Hz, which is unprecedented for terrestrial biological systems, and we did not have a priori notion about the locomotion dynamics at these high stride frequencies.

To what extent does the surface that the robot is operating on affect its performance, and can you elaborate on the potential surface attachment mechanisms" that you mention in the paper?

A good running surface for HAMR-Jr is one that is flat (low roughness), and has good friction. HAMR-Jr's vertical leg displacement is roughly 1 mm, and surface roughness of that order would be challenging. For surface features much larger, the robot would effectively need to climb" to overcome the obstacles. In the past, we have demonstrated that passive mechanisms like insect-like spines along the leg or gecko-like sticky pads can increase effective friction. We have also demonstrated active attachment mechanisms like electroadhesion that can be electrically modulated to stick to metallic surfaces like the inside of jet engines or even non-conductive substrates like the undersides of leaves. We are actively pursuing these as future research directions.

Photo: Kaushik Jayaram/University of Colorado Boulder/Harvard SEAS The researchers says that small robots like HAMR-Jr could help with applications in search and rescue, industrial inspection, environmental monitoring, and medicine.

What would be involved in developing an untethered version of the robot?

We have previously demonstrated power and control in HAMR-VI (as HAMR-F), which carries all the necessary onboard electronics in a ~1 g package along with a 0.33 g battery. As is, the same electronics can successfully power HAMR-Jr, which currently has a payload capacity of at least 3.5 g. However, we believe the electronics can be made even smaller. A recent paper from Robert Wood's lab showed that a ~89 mg electronics package could drive a 2-actuator robot like RoboBee. Replicating something similar, we may be able to obtain an electronics package (<400 mg) close to the weight of HAMR-Jr (~320 mg), without much optimization, which would be exciting!

Are there good reasons to make HAMR-Jr smaller, and how much smaller could HAMR robots get?

For a number of real world applications, we would like HAMR-Jr to be smaller and possibly more capable. Specifically, crevices in the rubble from collapsed structures are often just a few centimeters long. Similarly, in one of our previous collaborations with Rolls-Royce for engine inspection tasks, we found that typical boroscope port diameters are 8 to 12 mm in commercial jet engines. For robotic surgery related applications, the constraints are even smaller, as the largest arteries are only about 8 to 10 mm.

Making at-scale robots also makes it easy to use them as platforms for testing biology hypotheses (especially about insect locomotion) without worrying much about the physics of scaling, which is often unknown.

I hope to see a fully autonomous (both power and control) and highly capable version of HAMR-Jr fit within a one centimeter cube within the next few years.

What would you like to see robots like HAMR-Jr being used for?

The four main avenues I see small robots like HAMR-Jr make a positive societal impact in the next few years are in the domains of search and rescue (because of its mechanical dexterity), high value asset inspection (because of its high degree of autonomy), environmental monitoring (because of its scalability in terms of numbers) and medicine (because of its scalability in terms of size).

Scaling down an insect-size microrobot, HAMR-VI into HAMR-Jr," by Kaushik Jayaram, Jennifer Shum, Samantha Castellanos, E. Farrell Helbling, and Robert J. Wood from Harvard and UC Boulder, is being presented at ICRA 2020. < Back to IEEE Journal Watch