[Tom Ladyman] is making the case that a robot can take the place of a guide dog. According to his presentation, guide dogs cost about £45,000 (around $70k) to train and their working life is only about six years. On the other hand, he believes that this robot can be put into service for about £1,000 (around $1500). The target group for the robots is blind and visually impaired people. This makes since, because the robot lacks a dog’s ability to assist in other ways (locating and returning items to their companion, etc.). The main need here is independent travel.
He starts with the base of an electric wheelchair — a time-tested and economy-of-scale platform. The robot navigates based on images from four downward facing cameras mounted on the pole seen above. The X on the top of the pole allows for a much wider range of sight. The robot identifies its companion via a tag on their shoe, but it’s got another trick up its sleeve. The cameras feed to a set of four BeagleBoards which work together to process them into a 3D map at about 12 FPS, allowing for obstacle avoidance.
Check out the video after the break for a bit more information. The 3D guidance system is also explained in detail at the link above.
34 thoughts on “Can A Robot Be A Safe And Cost-effective Alternative To Guide Dogs?”
But can it love?
I had the crazy idea, of strapping a kinect , a pi, and a motorized pin thing, you know the ones you put on your hand and the pins move to make a 3d hand, If you motorize one, then you could make it form the environment the blind person is in. Rather then rely on the dogs senses encourage the blind persons tactile sense to take over for the loss of sight. It could potentially be used like brail, or even attached to the chest or back.
Totally possible, one blind guy clicks with his mouth and uses the sound like sonar to get around.
For just a second I thought it said “guard” instead of “guide” and got all excited!
Isn´t it problematic on stairs?
Stair, the arch-nemesis of the mechanical guide.
Why the “platform” – why not the sensors and some haptic feedback? Put on your “walking cap” and head out…so long as your legs still work it’s going to be the best solution – unless of course the blind person doesn’t like the way they look in a hat…. :/
I think it’s probably for two reasons (guessing at the design logic):
1) A dog also acts as a “sign post” that someone is blind – it’s clear from a distance to a normally sighted user that the dog user has a disability, and also the nature of the disability. This is very valuable facet, and one of the reasons hearing dogs are so successful.
2) Blind people already know how to interact with a dog and are familiar with the concept – the usability (wheel related issues notwithstanding) is higher in this form factor.
Using a wheelchair as a platform restricts able-bodied people to travel independently only where the wheelchair will go. I’d think a dog would have obvious advantages on stairways, escalators, curbs, uneven hiking trails, etc.
I truly applaud the noble use of technology to help the visually impaired. But I know quite a few guide dogs, and I applaud those noble dogs and those who train and work with them, as well. If I ever lose my vision and need assistance, as much as I love technology, I know I’d prefer a dog.
But keep working on and improving the robotics.
Sorry but guide dogs dont cost that much. Unless in the UK they have the royal family training them. Most Dogs here in the states are trained by volunteers.
I’ll rise to the troll, since you obviously know nothing of which you speak. Dogs are professionally trained. Typical pattern is 1 year residence with volunteer puppy trainer, who has to be closely supervised and reviewed by a professional trainer on an ongoing basis, then 3-6 months intensive training with a professional trainer. The 45k cost includes food, vet, and training costs – amortised over more than one dog, since there is only about a 70% success rate.
Oh come on now, that service animals, are trained by volunteers is largely accurate. No doubt that’s the case world wide, not just the US. That doesn’t mean that the volunteers are unprofessional, and even with unpaid professional trainers, getting service animal to those who need them isn’t expensive.
but can it poop? it would be funny if it dropped little presents from a ‘poop chute’
Well if it used disposable batteries, and had a self feed magazine of spares, it could poop out the expended ones.
It makes since?
The rest of that sentence states that the robot “…lacks a dog’s ability to assist in other ways…”. Unfortunately those ways are probably the things a blind person could use the most.
For example, there’s a blind lady that lives in my apartment complex. A few weeks ago while I was getting the mail, she was there, and she had unknowingly dropped one envelope from her stack of papers. Since I was there, and she has no dog, I said something and handed her the dropped piece. Any guide dog should be able to do the same. Let’s see a wheelchair-based robot do that. Let’s see the robot even notice that she dropped a piece of mail in the first place.
So can it cross the street, ie. know what traffic lights and don’t walk signs mean? If not I can see a lot of squashed blind people in the future.
Guide dogs don’t know that information. They know to identify the edge of the street and regular crossings, but the blind person still has to choose the safe moment to cross. That’s why ‘walk’ lights also make noise.
There’s no reason a decent machine can’t do the same as a guide dog, but they have three disadvantages at the moment:
-Offroading and uneven surfaces.
-Battery life. A guide dog requires one meal a day and will keep working without it.
-Memory. Guide dogs quickly memorise all your normal routes and will take you along them effortlessly.
These are all fixable, but there’s a few years left before we have something that can actually replace a dog.
My understanding guide dogs are trained to alert their owner if there’s a danger in crossing a street. Not all intersections are controlled by lights, and an audible walk signal doesn’t mean it’s safe to enter the crosswalk. Even if the dogs aren’t trained to do so, I imagine most dogs good enough to be quide dogs, will pick that up without being formally trained to do so.
For some guide dog facts. So many people assume all sorts of things about guide dogs, but, in reality, they avoid obstacles and not a huge amount else. Certainly nothing a robot couldn’t replace.
Fido Stay’s !
A valiant effort to solve a problem, but only one small facet of it has been explored. Helpling their owners walk down a (perfectly flat) street without hitting anything is probably a tiny percentage of what a guide dog does. Heck, if the problem is cost, use Pigs instead. Theyre cheaper to buy and feed for starters.
Pay thousands more for something that costs hundreds more to maintain and has less than half the capabilities even if properly implemented with existing planner algo.
I love ‘smart people’
A guide dog is much more than a walking obstacle avoider..
But assuming your looking for alternatives, surely it makes MORE sense to attach sensors to the blind person and provide feedback?
Think head mounted camera with small processor to detect hazards, I.E. walk signs etc…
Ultrasonics/Lidar with haptic feedback to give distances to objects..
there are alternatives already, but none have really been pushed to the blind community as much as the reliable guide dog.
Think comfortable sunglasses with a coin cell camera PCB with efficient planner and depth-mapping algo firmware, maybe two SMD proximity sensors. This is doable currently and is the most intelligent electrical solution
The problem is the firmware takes epic lab testing and turn-around, on top of the fact nobody would fund it cause of the limited market.
A talented coder would take a least a year to get the code to testing stage, and that’s in a high-level language; it’d probably take a FPGA and vlog-or-vhdl because the power source and clock-saving.
Of course I’m just passing time discussing this, I care as much about it as the people with funding and resources do..
It goes in the ‘things science would produce if humanity wasn’t self-righteous and greedy,,, book’
If no big money is available, kickstart it.
I’ll kickstart enzyme research for controlled cancer and human virus cures too, and wait for someone who made wealth off sports betting and/or industrial-piracy to see near-future marketing potential and give me funding -_-
The algorithms to do this right are established, it’s getting them to efficiently use clock cycles for usable cell-life that is the big production.
Algorithms actually solve a lot of human problems when embedded with censors, but the logistics are kamikaze to the life of a lone developer with little to no budget, and cause their is low-margin marketing for this it’s hard to get funding.
Sorry for the cynicism.. I guess it comes from working for IBM and Intel and being told to work toward a dull non-productive roadmap.. I see why stuff that remedies the human condition is what we should really be working on.. existing focused research solves nothing.
He’s absolutely right. Both service dogs and helper monkeys have been an inspiration for roboticists for many years. Back in 2007-2009, we did a bunch of research on this topic… you should check it out: Hizook.com overview and Lab Webpage.
Im allergic to dogs so if I ever became visually impaired I would almost certainly go for a robot – especially if it’s that cost effective. I agree that the finished product should be smaller etc but using the wheelchair as a prototype to prove the obstacle avoidance method using the cameras is very interesting!
Let’s go a “step” further! Put the chair back on the platform for more users.
Reasonable to expect that Tom has the best intentions in mind. IMO even if a robot can attain the t physical prowness of a dog it will always be second best to a dog. My concern is being less expensive that the disabled would be forced to settle for second best. In my mind the solution to the cost of service animals is to actively find out how to reduce the cost of getting them to those who need them. That’s such a no brainier, so that’s being done, but not enough people know about it so they can inquire as to how they can help.
No matter how good the technology is, it’s pretty charmless compared to a living, breathing dog. There are already plenty of phone apps that can be used to detect obstacles, which is far from being a total solution for blind people. A white cane does a lot of this, come to that, and at much less cost.
But to go for an entirely technologyical fix is to ignore the social and emotional sides of the whole life problem that many blind people face.
Being purely selfish, who is going to fund even the £1,000 that would be beyond the reach of many, given that most blind people are not at work, and the majority are in their pensionable years.
I can’t somehow see the public putting their hands in their pockets so readily for a Robots for the Blind Fund as they do unfailingly for the Guide Dog organisations. So this technology may be available to only a lucky few who have persuaded Social Services or Access to Work to fund it.
What kind of a companion would a robot be? Many blind people are isolated, or live vies that are fairly socially restricted. A dog not only can provide great companionship and entertainment, but can also make a blind person more of a colourful local figure in their community. This part of it might well put some blind people off the idea of having a dog, but having one can lead to a great deal more social interaction as you find your way about and exercise the dog.
A working life of 6 years for a dog strikes me as a mite pessimistic. My dog moved in at the age of 1 year, 11 months, and, all being well, I can expect him to work for me until he’s 9 or 10. Someone near me has a retired dog aged 15, so this does suggest how loyal dogs and their owners can be to one another for a very long time.
Just on practical mobility, what about all the “find” commands my dog knows – find the crossing, find the path, find the bin, find home….. I wouldn’t know the GPS co-ordinates for any of these things, but my dog knows what I mean, and can work out quite complex routes out of a jam. The dog is intelligent and can learn new things.
Every week, some gadget appears in my news feeds that claims to be a replacement for a gide dog, but it’s strange how none of them have actually replaced the dog.
Very good idea. Saw this guy in Nairobi working on the same exact thing. He is already conducting trials. His link is nairobi-uav.blogspot.com
The robot is clearly a prototype and there are some kinks to be ironed out before a final release. The method is interesting – it’s very different to other alternatives and could definitely be useful in the production of all sorts of robots – for example, house assistance robots.
Please be kind and respectful to help make the comments section excellent. (Comment Policy)