SHARE

Glidance CEO Amos Miller is one of an estimated 253 million people worldwide who live with a moderate to severe vision impairment. The overwhelming majority of those people currently navigate through the world without access to highly trained guide dogs or difficult to master walking canes. Miller, who was diagnosed with retinitis pigmentosa when he was five and has since lost most of his sight, believes his company’s autonomous guide robot could offer a solution to that mammoth accessibility gap. At just five pounds, Miller says “Glide” can spot obstacles and safely navigate blind people to their destinations, all at a fraction of the cost it takes to train and maintain highly specialized guide dogs. 

“In order to make a material difference to somebody who’s not very confident getting out and about, you need something that is physically connected to the ground and guide you,” Miller told PopSci. “And that’s what Glide is.”

“…we have these autonomous vehicles, we have self-guided drones, we land these drones on Mars and I’m sitting there at the airport and waiting for somebody to come to guide me to my gate. There’s something off here.”

Miller isn’t alone. Researchers across multiple continents are conducting experiments and testing the viability of robots, some of which happen to look like dogs themselves, as aides for the blind. If successful, these devices could bring an added layer of accessibility to large chunks of partially sighted and blind persons who’ve largely been left behind as autonomous technology has advanced around them. The Glide builds off of years of advances in robotics research and may represent one of the most promising new tools for aiding accessibility. Still, researchers say the technology isn’t poised to replace guide dogs anytime and would more likely attempt to fill in accessibility gaps for people who are unable or interested in owning a dog.  

How does the Glide robot work?

Though researchers have explored the idea of guide robots for at least five years, companies are just now on the verge of bringing products to market. Glidance, which was founded last year, is developing a robot ally called Glide, which it describes as the world’ “first self-guided primary mobility aid.” It resembles a small vacuum cleaner with a handle and two small wheels. The 9 by 9 inch robot uses “passive kinetic guidance” in place of a motor, so users simply push it forward to start it. Glidance says the device can be charged using a standard electrical outlet and can last up to 8 hours of “active use.”

“It’s as easy and familiar as holding onto someone’s hand.”

Glidance says the Glide uses an array of onboard sensors, cameras, and AI to analyze the immediate surroundings and guide users away from hazards. A haptic feedback sensor in the handle sends feedback to the users instructing them when they should slow down. Miller says the device will work with navigation apps which will allow users the ability to simply input a destination and have the device guide them towards it, not unlike an autonomous vehicle

Robots photo

This video shows legendary blind musician Stevie Wonder using Glide to walk through a room during this year’s Consumer Electronics Show in Las Vegas.

Glidance CEO Amos Miller, who lost most of his eyesight in his early twenties, told PopSci he came up with the idea for Glide device partly in response to personal mobility challenges he faced in his own life, especially during travel. Despite rapid advances in robotics and autonomous technology in other sectors like automobiles, the technology appeared to have left blind people behind. 

“There came this moment where you kind of say, okay, we have these autonomous vehicles, we have self-guided drones, we land these drones on Mars and I’m sitting there at the airport and waiting for somebody to come to guide me to my gate,” Miller said. “There’s something off here.” 

In the US, according to the guide dog breeder Guiding Eyes, only 2% of partially sighted and blind persons had guide dogs in 2017. Though these dogs are often provided for free, training and breeding them can cost upwards of $50,000 per dog, and only around half of those puppies pass their training. Working guide dogs are also typically retired and replaced after five or six years. All of those elements combined make the demand for these dogs far greater than the supply. Guide robots, in theory, could fill in part of that gap. 

Though past research has explored using Boston Dynamics’ Spot and other four-legged dog-like robots as guides, Miller believes those approaches won’t realistically work at scale. Aside from prohibitively high upfront cost, those large, bulky machines are also ill suited for mobile partially sighted and blind persons who need to move around the world. Unlike a large quadruped, Miller says users can easily collapse the Glide’s handle and take it with them on a bus or airplane. It’s also less likely to attract unwanted attention than a four legged robot imbued with sci-fi horror stereotypes. 

“People ask me, why doesn’t it have four legs and climb up steps and the reason is because it’s going to cost you $75,000 if it did,” Miller said. “If we are going to have impact at scale, we have to play the game, we have to leverage the incredible technology coming out of the autonomous vehicle industry but build a super simple device [that’s] powerful and affordable.”

Miller described the experience of using the Glide as similar to holding onto another person’s arm for guidance. When the device ships, Miller says users can operate in either a spontaneous or directive mode. The former is intended for users who know roughly where they want to go but who use the Glide to keep them on a straight path and avoid potentially hazardous obstacles. The directive mode, by contrast, will let users enter a specific destination and have Glide guide the user through haptic feedback and audio on how to get there. Miller gave the example of a traveler connecting Glide to an airport app which then instructs the device how to get to a gate. Users can also pre-program their own routes, which lets the Glide remember the directions to a local coffee shop or other frequent destinations. (Glide will utilize voice commands and accessibility options of existing apps.)

“You just walk and glide autonomously, Glide steers the way, avoids obstacles and gets you to your destination,” Miller said. “It’s as easy and familiar as holding onto someone’s hand.”

It’s unclear exactly when partially sighted and blind persons will get to that hand-holding sensation though. Glidance is taking preorders for the device and expects a beta for it to shop later this year. Miller says the device will likely have an unspecified up front cost as well as a subscription. That subscription price will vary depending on the level of features users want. Miller said Glidance plans to use a “Tesla model” where Glide ships with a baseline set of features and then improves via over the air updates over time.

Researchers are testing various types of guide robots 

Previous research into guidance robots have provided mixed results. One 2022 study conducted by a University of Copenhan researcher ran an experiment where a person walked through a path obstructed by a pallet, first with their guide dog, and again while holding onto Boston Dynamics Spot quadruped equipped with “sensing and semiotic capabilities.” The guide dog noticed the pallet and gently judged its owner to the side, avoiding the obstacle altogether. Spot, on the other hand, also noticed the pallet but instructed the blind person to walk on top and over it. Both the robot dog and the human stepped up the pallet in an awkward, misaligned, and potentially dangerous angle.

More recently, researchers from Binghamton University in New York used a reinforcement learning AI system to quickly train a four-legged robot to successfully guide a blind person around obstacles in a lab hallway. Computer Science Associate Professor Shiqi Zhang, who was involved with the project, told PopSci research into assistance robots has accelerated in recent years thanks to rapid advances in artificial intelligence and greater access to once wildly expensive robots. Still, Zhang agrees the hulking size of those robots likely made them ill-suited to assist blind people in everyday life. 

 “To serve the people with visual impairment, we need smaller size for charging, lower cost, and pretty easy to replace components,” Zhang said. “I think the goal is different and [Boston Dynamics’] Spot robots are not developed for this purpose. So there is a gap.”

Are robots better than guide dogs?

Miller stressed that he isn’t looking to replace guide dogs which he described as the current “gold standard” for getting around. Instead, he hopes the Glide device can cater to the vastly underserved 98% of partially sighted and blind persons who either can’t access a dog or don’t want the responsibility of owning one. But Miller said there are certain aspects of a robot guide that could make it more useful even than a well trained dog. Though guide dogs can learn some routes, that pales in comparison to on demand navigation tools available in a robot. And unlike guide dog owners, robot owners also don’t have to spend weeks training it or navigate around laborious travel restrictions. Robots also crucially don’t shed or defecate.

“You get the dog without the responsibility of the dog,” Miller said. 

“I’m not trying to compete with the dog. I’m trying to provide a solution for people who don’t use them.”

Zhang says robots also potentially have a leg up over dogs thanks to their upgradability. Researchers and product engineers can feasibly patch bugs and add new software quickly via updates. The robots could also be upgraded over time with new sensors or other hardware. In other words robots, unlike old dogs, can learn new tricks.

But that doesn’t mean it’s time to write-off dogs altogether. For now, guide dogs are much more adept at navigating their physical environment than their robot counterparts. Dogs can adjust to slippery terrain and are highly adept at spotting potentially dangerous situations. And then there’s the stair issue. Dogs have no problem paying their way to the second floor of an apartment building but the same can’t be said for robots. 

“Most dogs can do that [climbing stairs] without much difficulty, but it’s actually a pretty challenging problem for quadruped robots,” Zhang said. “Robots are able to do that, but it really requires a lot of computational resources, fancy sensors and real-time systems to get there.”

There’s also the intangible element of comparisonship dogs can provide, Zhang said. Some people simply like their dog and find comfort in their presence. While Miller said his device isn’t necessarily trying to compete with dogs on compassion, he claims he has already seen talk with the robot as if it were an “embodied agent.” Unlike a dog, the robot can also talk back. 

“Once we have voice interaction done I do think that there will be a sense of relationship,” Miller said.

But guide robots have skeptics as well. A spokesperson for the National Federation of the Blind told PopSci many past inventions have tried to take the place of tried and true canes and guide dogs but had failed. Jason McKee, an instructor at Guide Dogs Queensland in Australia recently told the ABC robotic dogs were impressive machinery and could be useful for traveling internationally but still couldn’t compare to a well trained canine. 

“Nothing will beat the companionship, the intuition of the dog, then be able to forward think, see ahead and then make the movements to guide someone safely,” McKee said. 

Miller, on the other hand, believes there’s a future where guide dogs and robots exist in harmony, complementing each other rather than competing. 

“I want to be very clear that my goal is to empower people who are not using dogs or cans today. I’m not trying to compete with the dog,” Miller said. “I’m trying to provide a solution for people who don’t use them.”

Update 03/12/24 2:44 PM: The weight of Glide has been updated. A voice command feature mention has been added.