Every object is the results of an experiment in robotic autonomy. By itself, studying because it goes, the robotic is trying to find, and making an attempt to make, an object with essentially the most environment friendly energy-absorbing form to ever exist.

To do that, the robotic creates a small plastic construction with a 3D printer, data its form and dimension, strikes it to a flat metallic floor — after which crushes it with a stress equal to an grownup Arabian horse standing on 1 / 4. The robotic then measures how a lot power the construction absorbed, how its form modified after being squashed, and data each element in an unlimited database. Then, it drops the crushed object into the field and wipes the metallic plate clear, able to print and take a look at the following piece. Will probably be ever-so-slightly completely different from its predecessor, its design and dimensions tweaked by the robotic’s laptop algorithm based mostly on all previous experiments — the premise of what is referred to as Bayesian optimization. Experiment after experiment, the 3D constructions get higher at absorbing the impression of getting crushed.

These experiments are potential due to the work of Keith Brown, an ENG affiliate professor of mechanical engineering, and his workforce within the KABlab. The robotic, named MAMA BEAR — brief for its prolonged full title, Mechanics of Additively Manufactured Architectures Bayesian Experimental Autonomous Researcher — has developed because it was first conceptualized by Brown and his lab in 2018. By 2021, the lab had set the machine on its quest to make a form that absorbs essentially the most power, a property generally known as its mechanical power absorption effectivity. This present iteration has run repeatedly for over three years, filling dozens of packing containers with greater than 25,000 3D-printed constructions.

Why so many shapes? There are numerous makes use of for one thing that may effectively take in power — say, cushioning for delicate electronics being shipped internationally or for knee pads and wrist guards for athletes. “You can draw from this library of information to make higher bumpers in a automotive, or packaging gear, for instance,” Brown says.

To work ideally, the constructions must strike the right stability: they can not be so sturdy that they trigger harm to no matter they’re supposed to guard, however needs to be sturdy sufficient to soak up impression. Earlier than MAMA BEAR, the most effective construction anybody ever noticed was about 71 p.c environment friendly at absorbing power, says Brown. However on a cold January afternoon in 2023, Brown’s lab watched their robotic hit 75 p.c effectivity, breaking the identified document. The outcomes have simply been revealed in Nature Communications.

“After we began out, we did not know if there was going to be this record-breaking form,” says Kelsey Snapp (ENG’25), a PhD scholar in Brown’s lab who oversees MAMA BEAR. “Slowly however certainly we stored inching up, and broke by way of.”

The record-breaking construction seems to be like nothing the researchers would have anticipated: it has 4 factors, formed like skinny flower petals, and is taller and narrower than the early designs.

“We’re excited that there is a lot mechanical information right here, that we’re utilizing this to be taught classes about design extra typically,” Brown says.

Their intensive information is already getting its first real-life software, serving to to tell the design of recent helmet padding for US Military troopers. Brown, Snapp, and venture collaborator Emily Whiting, a BU Faculty of Arts & Sciences affiliate professor of laptop science, labored with the US Military and went by way of area testing to make sure helmets utilizing their patent-pending padding are comfy and supply ample safety from impression. The 3D construction used for the padding is completely different from the record-breaking piece — with a softer heart and shorter stature to assist with consolation.

MAMA BEAR is just not Brown’s solely autonomous analysis robotic. His lab has different “BEAR” robots performing completely different duties — just like the nano BEAR, which research the way in which supplies behave on the molecular scale utilizing a expertise referred to as atomic drive microscopy. Brown has additionally been working with Jörg Werner, an ENG assistant professor of mechanical engineering, to develop one other system, generally known as the PANDA — brief for Polymer Evaluation and Discovery Array — BEAR to check 1000’s of skinny polymer supplies to seek out one which works finest in a battery.

“They’re all robots that do analysis,” Brown says. “The philosophy is that they are utilizing machine studying along with automation to assist us do analysis a lot sooner.”

“Not simply sooner,” provides Snapp. “You are able to do belongings you could not usually do. We will attain a construction or objective that we would not have been capable of obtain in any other case, as a result of it might have been too costly and time-consuming.” He has labored intently with MAMA BEAR for the reason that experiments started in 2021, and gave the robotic its skill to see — generally known as machine imaginative and prescient — and clear its personal take a look at plate.

The KABlab is hoping to additional exhibit the significance of autonomous analysis. Brown desires to maintain collaborating with scientists in varied fields who want to check extremely giant numbers of constructions and options. Regardless that they already broke a document, “we’ve no skill to know if we have reached the utmost effectivity,” Brown says, which means they might probably break it once more. So, MAMA BEAR will carry on operating, pushing boundaries additional, whereas Brown and his workforce see what different purposes the database may be helpful for. They’re additionally exploring how the greater than 25,000 crushed items may be unwound and reloaded into the 3D printers so the fabric may be recycled for extra experiments.

“We will preserve finding out this technique, as a result of mechanical effectivity, like so many different materials properties, is barely precisely measured by experiment,” Brown says, “and utilizing self-driving labs helps us decide the most effective experiments and carry out them as quick as potential.”