Today we’re excited to announce that we’ll be joined by autonomous vehicle pioneer Anthony Levandowski.
Levandowski is the co-founder and CEO of Pronto, a new startup developing aftermarket kits for semi-trucks. The company is touting the safety potential of autonomous vehicle technologies for cross-country hauls. “We are not building technology that tells vehicles how to drive,” he wrote in a Medium post announcing Pronto’s launch. “Instead, our team of engineers is building tech that can learn how to drive the way people do.”
Before founding Pronto, Levandowski served as the head of Uber’s autonomous vehicle program and helped Google develop its own initiative that ultimately became Waymo. The engineer began his career by building the autonomous motorcycle Ghostrider for DARPA’s Grand Challenge.
If you’ve ever dabbled in collectible card games — Magic: The Gathering, Pokémon, etc. — you know how quickly collections can grow. One pack turns into two. Two turns into five. Then they release some new set and… screw it, why not buy a whole box?
Card resellers have the same problem, just magnified to an extreme. People who’ve stopped playing a game for whatever reason (sometimes years prior) walk in with massive collections and just want to get rid of them. Online resellers and card stores can end up with monstrous stockpiles of unsorted cards, and going through them requires a ton of time and a wealth of ultra-specific knowledge of a game. Which cards are rare? Which ones are a bit more common, but useful enough that players would want to buy them for their decks? What are they all worth?
Sorting Robotics, a company in Y Combinator’s Winter 2019 class, has built a robot laser-focused on that problem. You load it with up to 1,000 Magic cards, and it’ll automatically sort them to your liking, look up their values and give you all the data in a big spreadsheet.
The machine is able to sort by a bunch of different criteria, be it alphabetically, by the set a card is from or by its resale value (as pulled from TCGPlayer). Want a big pile of all of the cards worth more than $1? It can do that. If you need them sorted other ways, the company is open to helping with custom sorting logic.
One challenge the team had to tackle early on was how to handle cards with minimal contact for the sake of preventing possible damage. Some Magic cards, after all, resell for hundreds or thousands of dollars — if their machine got a reputation for damaging cards even occasionally, no one would use it.
So Sorting came up with a pneumatic system that uses cameras, computer vision, polished surfaces and silicone suction cups to identify and move cards from stack to stack with limited contact. There are a few fancy tricks involved, like picking up cards in a way that utilizes the airflow within the machine to keep it from lifting two lightly stuck together cards at once. If a card is loaded into the machine upside down, it’ll shift it into a pile with other upside down cards to be manually flipped and re-sorted later. Sorting 1,000 cards takes 1-2 hours, depending on the criteria by which they’re being sorted.
If a card does somehow get damaged, Sorting Robotics says they’ll cover the cost. (They’ll want to check the feed from a pair of cameras inside the machine to see exactly what happened, so you probably shouldn’t go throwing an already bent up Time Walk card in there and asking for reimbursement.)
Another challenge: dirt. Even for collectors, cards are rarely 100 percent pristine. There’s the natural oils from your hand, the dust from being stored over time and even some amount of the card’s own dust, left over from the printing and cutting process. You might not really notice it if you’re just dealing with your own collection — but when you’re putting thousands of cards through a machine with moving parts and camera lenses, the dust adds up fast. Later versions of their machine have been re-tailored to better deal with dust, and to be more easily maintained when the dust builds up.
Sorting has three founders: Nohtal Partansky and Sean Lawlor (both of whom were previously systems engineers at NASA’s Jet Propulsion lab), and Cassio Elias dos Santos Junior, a computer vision engineer who previously built a popular Magic card scanning app for Android.
As for how much it costs, the company would only say that it’s working on that on a case-by-case, shop-by-shop basis. They stressed that they’re focusing on building these for online resellers and card shops — so it sounds like it’s not in the price range that most hobbyists might consider.
The machine currently only sorts Magic cards, though the founders tell me support for Yu-Gi-Oh! and Pokémon cards is coming shortly.
It’s already tomorrow in Australia, seemingly in more ways than one. It’s the 27th already, yes, but they’re also working in putting together AI-flown companion jets for their fighters. Why didn’t we think of that? It’s a Boeing Australia joint, but maybe they’ll contract out to the U.S. facilities and we can snake one off the line. I know a guy.
The aircraft, currently in development but scheduled for first flight in 2020, is meant to be a loyal wingman to pilots flying military missions — as you might guess from its name, the “Loyal Wingman.” The official full name is the Boeing Airpower Teaming System,” which acronyms to BATS, but they don’t look or act much like bats so this probably won’t be emphasized.
Essentially these are drones that will accompany other craft, flying in formation and providing defensive capabilities. It’s a force multiplier, which is important for governments that can’t field as many pilots or primary craft (i.e. modern fighters) as countries that have invested more heavily in their air force.
Boeing International’s president, Marc Allen, emphasized (naturally) this international-enablement aspect of the craft in a statement:
This aircraft is a historic endeavor for Boeing. Not only is it developed outside the United States, it is also designed so that our global customers can integrate local content to meet their country-specific requirements. The Boeing Airpower Teaming System provides a transformational capability in terms of defense, and our customers – led by Australia – effectively become partners on the program with the ability to grow their own sovereign capabilities to support it, including a high-tech workforce.
In other words, it’s nice to see some investment outside the U.S., diversifying the portfolio a bit.
A full-scale mock-up was revealed at the Australian National Airshow today:
The Loyal Wingman is 38 feet long and should have a 2,300-mile range. It will fly independently but will almost certainly remotely as well, and can be equipped with a variety of sensor packages and other goodies. I wouldn’t expect these to get into any dogfights, however. They’re meant to be support, providing recon and surveillance duties that can’t be done from, say, a research or cargo craft.
Given the popularity, in military circles anyway, of drones as solo recon, this kind of “extra pair of eyes” duty makes a lot of sense, and seems inevitable. Whether Boeing’s approach will be the one to take off in governments around the world surely depends on the execution, so we’ll revisit this story in 2020 when the Wingman actually takes flight.
Researchers have created a prosthetic hand that offers its users the ability to feel where it is and how the fingers are positioned — a sense known as proprioception. The headline may be in jest, but the advance is real and may help amputees more effectively and naturally use their prostheses.
Prosthesis rejection is a real problem for amputees, and many choose to simply live without these devices, electronic or mechanical, as they can complicate as much as they simplify. Part of that is the simple fact that, unlike their natural limbs, artificial ones have no real sensation — or if there is any, it’s nowhere near the level someone had before.
Touch and temperature detection are important, of course, but what’s even more critical to ordinary use is simply knowing where your limb is and what it’s doing. If you close your eyes, you can tell where each digit is, how many you’re holding up, whether they’re gripping a small or large object and so on. That’s currently impossible with a prosthesis, even one that’s been integrated with the nervous system to provide feedback — meaning users have to watch what they’re doing at all times. (That is, if the arm isn’t watching for you.)
This prosthesis, built by Swiss, Italian and German neurologists and engineers, is described in a recent issue of Science Robotics. It takes the existing concept of sending touch information to the brain through electrodes patched into the nerves of the arm, and adapts it to provide real-time proprioceptive feedback.
“Our study shows that sensory substitution based on intraneural stimulation can deliver both position feedback and tactile feedback simultaneously and in real time. The brain has no problem combining this information, and patients can process both types in real time with excellent results,” explained Silvestro Micera, of the École Polytechnique Fédérale de Lausanne, in a news release.
It’s been the work of a decade to engineer and demonstrate this possibility, which could be of enormous benefit. Having a natural, intuitive understanding of the position of your hand, arm or leg would likely make prostheses much more useful and comfortable for their users.
Essentially the robotic hand relays its telemetry to the brain through the nerve pathways that would normally be bringing touch to that area. Unfortunately it’s rather difficult to actually recreate the proprioceptive pathways, so the team used what’s called sensory substitution instead. This uses other pathways, like ordinary touch, as ways to present different sense modalities.
(Diagram modified from original to better fit, and to remove some rather bloody imagery.)
A simple example would be a machine that touched your arm in a different location depending on where your hand is. In the case of this research it’s much finer, but still essentially presenting position data as touch data. It sounds weird, but our brains are actually really good at adapting to this kind of thing.
As evidence, witness that after some training two amputees using the system were able to tell the difference between four differently shaped objects being grasped, with their eyes closed, with 75 percent accuracy. Chance would be 25 percent, of course, meaning the sensation of holding objects of different sizes came through loud and clear — clear enough for a prototype, anyway. Amazingly, the team was able to add actual touch feedback to the existing pathways and the users were not overly confused by it. So there’s precedent now for multi-modal sensory feedback from an artificial limb.
The study has well-defined limitations, such as the number and type of fingers it was able to relay information from, and the granularity and type of that data. And the “installation” process is still very invasive. But it’s pioneering work nevertheless: this type of research is very iterative and global, progressing by small steps until, all of a sudden, prosthetics as a science has made huge strides. And the people who use prosthetic limbs will be making strides, as well.
Manmeet Maggu and Rahul Udasi didn’t know it when they met at the University of Waterloo eleven years ago, but the bond that the two forged in late night study sessions as roommates in the UW Place dorm has helped power their work building an exoskeleton that allows children with disabilities to walk.
The fruit of that labor is Trexo Robotics, which will graduate as part of the latest batch of Y Combinator’s winter 2019 cohort.
In the three years since Udasi and Maggu launched their company, Trexo has gone through the Techstars accelerator program in New York and raised $720,000 in seed funding. But the roots of the company extend back to Maggu’s senior year at Waterloo, when he learned that his nephew in India had been diagnosed with cerebral palsy.
The disease, which affects millions around the world and at least 500,000 children in the U.S. alone, was not something that Maggu had known much about. But with the diagnosis of his nephew, he began to learn.
“We started looking at what cerebral palsy is and what it would mean to him,” says Maggu. “For him it meant that he would spend his entire life in the wheelchair and I knew the tremendous negative health effects associated with sitting.”
At first, the family looked at ways to encourage the child to walk outside of physical therapy, with no results.
“Initially we were like, let’s buy him a robotics system or an exoskeleton, but after looking around we saw that there was nothing out there,” says Maggu.
That’s when Maggu determined that he would make the development of the exoskeleton the focus of his senior design thesis, back in 2012.
Trexo co-founders Manmeet Maggu and Rahul Udasi (Image courtesy of Trexo Robotics)
“After that everybody went our own ways but I couldn’t stop working on this,” Maggu recalls. “I kept working on this on the side. I was just working in my apartment doing design, doing 3D printing there.”
By that time, Maggu and Udasi had gone their separate ways. Maggu worked on the project on the side while pursuing his career in the technology industry. He worked at Blackberry and Qualcomm, but kept in touch with his college friend, Udasi, had worked for a spell at Willow Garage and a few other robotics companies, before returning to Canada to study for a masters degree in robotics at the University of Toronto.
Maggu had also returned to Toronto to work on a masters in business administration.
But throughout that time, using a low-end printer that he bought for $600, Maggu kept prototyping. Then, when the two were living together, Maggu would ask Udasi, the robotics expert, for his help.
By 2017 the two men had developed a functioning prototype and flown to India to give it to their first test patient — Maggu’s nephew.
“The first time we tried it it didn’t work,” says Maggu. “But my brother has a factory in India in Delhi, so we made some more modifications and tried it out again and I watched my nephew try to walk with the device for the first time.”
After progressing through the Techstars accelerator and raising its seed round, Trexo now has around six versions of its exoskeletons in private homes and hospitals. The company is conducting a clinical study at the Cincinnati Children’s Hospital and has four paying customers in Canada.
These are still devices that aren’t affordable for most Americans — not by a longshot. Trexo is pursuing a direct-to-consumer approach that would see their technology selling for $899 per month through a lease model with a $1,000 down payment and financing for 36 months. Customers can also lease the product for $999 with at least a required 12 month lease period or they can buy a Trexo exoskeleton for $29,900.
The company is marketing the device as an exercise and therapy device, which means that it can avoid some of the regulatory requirements to bring the product to market under the current regulations from the Food and Drug Administration, which oversees medical devices.
Maggu and Udasi ultimately see Trexo tackling more than just treatment for children with mobility issues. Indeed, Maggu sees opportunities for the company to begin developing products for elder care as well.
“We view this as much more of a consumer product,” says Maggu. “We strongly believe that wearable robotic systems are going to play a huge role in the future to come.”