Nissan's Crash-Free R&D: 7 Cute Robots Mimicking Bees and Fish 105
cartechboy writes "As Nissan develops autonomous cars for its 2020 target date, the company's engineers are modeling the tech after behaviors seen in bumblebees and fish. Nissan actually tests self-navigation algorithms in seven small toy-looking robots called EPORO. The robots have 180-degree vision (modeled after bees) and monitor each others' positions, travel nose to nose and avoid collisions--just like a school of fish. Getting small robots to zip around without bumping into things might be the first step in getting cars to do the same."
travel nose to nose? (Score:3, Informative)
Fish don't travel nose to nose. That would cause issues. Nose to tail seems to work much better for them
Re: (Score:2)
I suspect they mean horizontally, not vertically. But that would be amusing, sort of a Fred & Ginger car configuration.
FINALLY! (Score:2)
Re: (Score:2)
If you remember that recent /. story about Asimov's predictions for the 2014 World's Fair, he went a bit far in predicting cars that, while the couldn't fly, could hover briefly on compressed air to cross obstacles.
However, he specifically predicted small robots darting through the crowd, avoiding accidents, as a tech demo for coming self-driving cars. Right on target it seems!
Really, the only reason we don't all have ground-effect hovercars is that power didn't get cheap at the expected rate, and we can't
Re: (Score:2)
And anyway, what obstacles do you encounter when you drive that would be useful to float over?
Priuses. If I could just hover across the cloud of smug, that would be bliss.
Re: (Score:2)
The main problem with flying cars is the power they use.... keeping stuff above the earth requires too much energy... they're possible now, but not really desirable.
Re: (Score:3)
Fish don't travel nose to nose. That would cause issues. Nose to tail seems to work much better for them
Fish don't travel nose to tail, either. They are staggered. Until we allow staggered "lanes" instead of rigid, fixed lanes, then we won't be traveling like fish.
Note also that fish are narrow, i.e. they have a small cross section for their volume. This helps to travel staggered. Maybe staggered-driving cars will need to be narrower, longer, and taller. Will people get used to one-across seating?
Re: (Score:1)
Fish also don't use roads, unless you live near Seattle, where they do.
Re: (Score:1)
The reason that fish travel staggered is the same as birds. They're using fluid dynamics to propel themselves. Cars by comparison are only slowed by fluid dynamics (unless you're talking about very high end sports cars), and hence would want to travel nose to tail (to stay in the slip stream) as much as possible, except on very windy roads, with particular car design characteristics.
Re: (Score:2)
Re: (Score:2)
Summary is wrong.... (Score:2)
I noticed two things right away when I read TFA:
1. The article has them traveling nose to tail- not nose to nose.
2. According to TFA, the robots have 180 degree vision, not the summary stated 360.
iRobots... (Score:1)
Does Apple already have a claim on the name? If Japan continues with this trend of cute robots, it'll be the next big thing.
Is Microsoft paying any attention to this?
Re: (Score:2)
Too late, iRobot makes vacuum cleaners. ;-)
But MSFT might sue them to get the name with some flimsy IP claim if they wanted it.
Re: (Score:1)
iRobot has a claim on the name iRobot.
Lurking Skynet approves (Score:5, Interesting)
Seriously, not a whole lot of zipping - more like ambling. I think they're going to need a whole lot better processing to handle movement at 45mph, much less 75mph.
Still, emergent behavior is definitely a strong idea... just wondering how the "groups" form - what sort of negotiation is needed? Will it require some form of authorization/authentication? What happens when the "group" loses an individual (ie, power/comm failure)? What about rogue elements?
Lots of stuff to study and apply - but it still looks far off in implementation. I'd love to see this research combined with Google's driverless car tech :)
Re: (Score:2)
It's swarm behavior, each unit of the group has simple rules that follows. Things like "don't get closer than X to the unit in front of you" and "don't approach another unit at more than Y relative speed". There's no central processing happening to manage the group, no more than there is for a flock of starlings flying through the air. It's just simple rules leading to seemingly complex behavior.
Re: (Score:2)
It's swarm behavior, each unit of the group has simple rules that follows.
It's not just a swarm - a swarm usually deals with members from the same hive or have some basis of affinity. How will we generate affinity between units securely? Auth?
For cars, the "swarms" will need to be created and modified ad-hoc (e.g.: car coming into right lane in highway from onramp, exiting, switching lanes, etc). That aspect of this seems very complex and the system needs to add in redundancy and robustness... not to mention, is there an analogue to ad-hoc swarms in nature?
They should make it resemble Bumblebee (Score:3, Funny)
and transform into a 30 foot tall robot with plasma cannons....
Seems a stretch (Score:2)
Re:Seems a stretch (Score:4, Insightful)
not all cars have to be autonomous for this to work. The autonomous cars will just be traveling in groups, which will be very efficient. I can even foresee dedicated lanes for autonomous cars.
Re: (Score:2)
Autonomous cars would have to cope with human driven cars... in which case, they'd be pretty easy to troll. I know I would :P.
Re: (Score:2)
They should throw some puppies in that racetrack, to simulate human drivers.
To simulate city driving, they could then tie a chew toy on the back of each robot.
Re:Seems a stretch (Score:4, Interesting)
I actually recently discovered why BMW drivers love your arse so much.
It's because of their active cruise control system. I was demonstrated a 5 series' cruise control system, which got so close to the car in front it terrified me. If I were driving the car in front, I would probably have been thinking "bloody beemer driver" at that point.
So actually, it's already the fault of autonomous cars that people get up close and personal behind you.
Re: (Score:2)
"You can set the specified minimum distance for DISTRONIC PLUS by varying the time span between one and two seconds. With this function, you can set the minimum distance that DISTRONIC PLUS keeps to the vehicle in front, dependent on vehicle speed. You can see this distance in the multifunction display
WARNING
It is up to the driver to exercise discretion to select the appropriate setting given road conditions, traffic, driver's preferred driving style and applicable laws and d
Re: (Score:2)
It seems that BMW has merely automated an activity that was previously done manually. Hopefully the cruise control system is less likely to be distracted by cell phone/latte/other shiny cars/etc. than the BMW drivers I typically see.
Re: (Score:3)
I don't get this... I don't let anyone sit on my arse, ever. If someone does get too close, in the overtaking lane, when I'm looking to overtake someone in a queue, I pull over. It's their problem, and I'm not going to be a part of it. I only sit on the arse of others when they do not pull over for no reason (actually, generally I undertake them, carefully).
I don't drive _that_ fast... generally about 80 when cruising in the UK, which is technically illegal, but what loads of other people do, and polic
Re: (Score:2)
There seem to be a lot of drivers who don't want to go fast, they just want to be first. They will go as fast as they need to to get in front of everybody, and then slow down. In the passing lane. Driving erratically like this seems much more dangerous than just speeding.
Re: (Score:2)
This seems like it utilizes swarm/schooling behavior.
Right. That goes back to Craig Reynolds' "Boids" paper, and is the basis for much crowd behavior in movies and video games. In the real world, it's less useful. The general idea is that there are attracting and repelling fields, and you add up the fields and get a direction vector. This works OK in not-too-crowded spaces. It's great for birds and fish. When there are turning circle limits, or narrow lane limits, it's not as useful.
We tried this approach on our DARPA Grand Challenge vehicle in 2004-2005. It
Zipping around? (Score:1)
Do you know how I know you didn't watch the video?
Little robots lurching and stumbling around at random does not impress me as a future smart-car building block.
Re: (Score:2)
Re: (Score:1)
My favorite example is driving in Seattle. Watching cars spin out as they slide down an icy hill, where most people don't turn into the spin, and then smash each other, or during a sudden downpour while your car is misted up and your sensors are shot - these are a few of my favorite things.
Zip my (axle).
Re: (Score:1)
The problem with your examples is that it is already proven that humans can't handle those situations satisfactorily.
I'm not even convinced that a self driving car with a few shot sensors will perform worse than most humans during those circumstances.
That depends on what you mean by satisfactorily.
To whom?
To the insurer? To the person? To the car?
If the car hits a house is this better or worse than hitting a puppy?
What if it hits a little kid building a snowman?
Re: (Score:1)
Um, not silly examples. Real world examples from the city I live in.
Look, what works in the lab does not always work in the wild.
a wet blanket (Score:5, Interesting)
"Getting small robots to zip around without bumping into things might be the first step in getting cars to do the same."
I seriously hope they are past the 'first step' of modeling things in small robots. If they are planning on releasing this thing on the road in the next six years, they need to have tech that is just being refined at this point.
For comparison, it can take six years to test and refine avionics software, even after all the algorithms are known. This software needs to be extremely reliable. Remember that even if a server has 99.999% uptime, it's still going to crash every year or so. When people's lives are on the line, you're going to want 99.99999% availability. That kind of software is not easy to make. If they are still doing fundamental research, they aren't going to have it done in time.
Re:a wet blanket (Score:4, Interesting)
Human drivers are far short of 99.999999% reliable, so I say hurry it up even if they're at Five Eights reliability...
Re:a wet blanket (Score:4, Insightful)
If you want to make a human comparison, you have to ask, "how often does a human have a heart attack or other catastrophic failure on the road?" This is just the base system, you need to make it reliable before you even get to talking about the quality of the algorithms. It's great you have a perfect driving algorithm, but if the OS crashes and needs to reboot once a year while driving, then no one will care if you have a perfect algorithm. Your car is having the equivalent of a heart-attack every year on the road.
It's harder than writing a typical website. To get an idea of the difficulties involved, remember that at that level of reliability you need to take into consideration that cosmic rays will corrupt your memory. ECC can be helpful, but you can still get corruption. So imagine if you have a for loop, and your index gets corrupted. How will you guarantee that it doesn't become an infinite loop?
These problems are known and can be solved, but the point is, it's really hard. It will take six years of effort, and that is assuming the algorithms are ready now. If they are not, then they won't have it ready by 2020.
Re: (Score:1)
Re: (Score:2)
He was asking the right question. Self-driving cars don't need to be much less mistake-prone than humans to be accepted. Crashing once per year would be a bit much, but as long as it crashes less frequently than the human would it's good enough for early adopters. The perfect is the enemy of the "good enough to sell" in this case.
Re: (Score:2)
Crashing once per year would be a bit much, but as long as it crashes less frequently than the human would it's good enough for early adopters.
The topic under discussion is the difficulty of creating a system that crashes that infrequently.
Re: (Score:2)
In which case talking about things like cosmic ray strikes flipping bits is moronic. That sort of thing is far under the noise floor. We simply don't need airliner levels of reliability to be acceptably better than human drivers.
Re: (Score:2)
Re: (Score:2)
Plus the average airplane gets many many more preventative maintenance hours performed on it than the typical car. Military planes more than civil, but still, MH per FH is significantly greater than 1:1 for aircraft, whereas for cars its significantly less.
Now for cars, that makes sense, or else the operating costs would skyrocket, so low maintenance needs are an essential feature. But then that means the parts and the software needs to be EVEN MORE reliable. And this is without even considering how poorly
Re: why robot cars dream of electric sheep (Score:1)
Actually, electric vehicles have a much lower maintenance requirement than gas/diesel vehicles do - on average about 1/4 the cost factor and time factor.
Requiring auto-guided vehicles be electric only, with max speeds of say 30 mph (typical highway speeds during rush hour) might go a long way to making them safe. Potential energy drops as speed drops.
Re: (Score:2)
Actually, electric vehicles have a much lower maintenance requirement than gas/diesel vehicles do - on average about 1/4 the cost factor and time factor.
The origin of the engine's power is irrelevant, unless somehow a car that is propelled by an electric motor is safer than the one that is propelled directly by an ICE.
Maintenance would be relevant in the areas of power {brakes,steering,etc.} but that is already electric in hybrids.
Requiring auto-guided vehicles be electric only, with max speeds of sa
Re: (Score:1)
The origin of the power train is relevant - fewer moving parts, less subject to corrosion, no controlled explosions (in grade 10 I took power mechanics, I could build an engine for you if you want.
Brake maintenance tends to be about the same.
Re: (Score:2)
that's great. but thats not my point. my point is how little maintenence the owners of the cars will actually perform. electrics have lower requirements? that's great, and comepletely beside the point that cars get treated like sht by a lot of their owners. yet the cars still have to have X level of reliability precisely because of that in order to meet all the regulations and be attractive to customers.
Re: (Score:1)
According to industry standards, total cost of maintenance is much higher for gasoline or diesel engine cars than for electric cars.
(source CNN Money)
Re: (Score:2)
Re: (Score:2)
First generation autonomous vehicles won't be able to use swarm logic, because it only works if everyone is doing it and 99% of the cars on the road won't be (they're be driven by people). Now, in 2040 when you simply can't buy a car that doesn't drive itself you can start looking at eliminating the "rules of the road" as they exist now. If all vehicles perform with near perfect decision making you can do things like divide the 2 lanes each way highway into 6 lanes (because they can be much closer togethe
Re: (Score:2)
Good point.
All research currently is with these small bots, minipucks and such. But research as also shown that mass doesn't scale well in the algorithms used for collision avoidance and path planning. They just are not robust enough to handle non-linearity that occurs in the environment--which means that demo will not scale well.
Why? What fun is an autonomous car? (Score:1)
Re: (Score:2, Insightful)
Every driven I-10 across Texas? Try it and you'll understand why you might want an autonomous car.
Re: (Score:3)
I-10 is a summer blockbuster movie compared to US-50 across Nevada.
What fun is a normal car? (Score:3)
What fun is a normal car? I mean, I can't think of any activity that involves nearly as much banal, repetitive tedium combined with the need to be vigilant against life-threatening danger that doesn't involve enlisting in the military.
I hate driving. You can't do anything really exciting with a car 95% of the time, because there are (necessary) safety laws preventing it, and the road is full of drivers worse than you (or at least worse then you think you are). I waste about 5% of my life every day on dri
Re: (Score:3)
Driving is fun. Traffic is boring. Driving without traffic is increasingly difficult to find. But there are plenty of corners on my daily commute to have fun with under 30 MPH. It doesn't have to be fast to be "as fast as you can".
Re: (Score:2)
They are seriously over estimating the demand for such vehicles...
I doubt it. It's not like it will debut as an expensive mandatory package on every car. It will be an (expensive) option on cars where it makes sense. Nissan just introduced steer-by-wire. They already have smart cruise control, lane drift prevention, blind spot detection, impact reduction, and so on, but the cars only steer themselves gently by using the brakes on one side.
In a few years I expect options that will give the computer far more control of steering, brake, and throttle in situations where t
Re: (Score:1)
I drive 20,000-30,000 miles a year here in the UK. 95% of that is on motorways in the leftmost lane (rightmost in your country) at 65mph bored shitless daydreaming. 20,000 miles at 65mph is 310 hours, which translates into almost 13 *full* days. 30,000 miles is 460 hours, more than 19 full days.
In my lifetime, I never have brought any car more than $5,000 and never will due to the mileage I drive causing steep deprecation[1], but I will buy the first reasonably priced car that has this tech brand new eve
This is the way to go. (Score:5, Interesting)
This kind of system needs to be based on natural and fluid situations. Trying to base them on as-presently-constituted traffic laws is a mistake no matter how you slice it. The paradigm has different advantages and shortcomings than manual driving. Build a good anti-collision system, and then as needed, add the other layers on top instead of building a base upon assumptions based on law.
Re: (Score:2)
This kind of system needs to be based on natural and fluid situations.
If the autonomous cars can ignore accidents on the other side of the highway I'm all for it.
Re: (Score:2)
Current laws are based on the constrains of roads while bees and fish work in a different world. The most obvious being 3 dimensional for the natural system vs 2 dimensional for most roads. There is also the constrains of the sides of the roads, intersections, on/off ramps, different vehicle performances, higher speeds, greater momentum, greater consequences of collision,etc. Natural situations are much more forgiving to errors. One fish hitting another will cause little or no damage. One car hitting anothe
Re: (Score:1)
Yes. Which is why you build the natural system and then build those considerations as a restriction package on top of it. There is no reason to build a system that is dependent on the status quo, all that does is impede future improvements.
Re: (Score:2)
If you build a natural system that can not fit into the constrains you are no further ahead and have just wasted lots of time and money. For example, bees and fish can go over/under each other if crossing; vehicles can not. Failing to take that constraint into account in the original design will doom it to failure. with all these systems the 80/20 rule applies. 80% of the work will be consumed by the last 20% of the problems.
There is no reason to build a system that is dependent on the status quo,
There is a reason and that is due to the fact that we all are not going to instant
Re: (Score:1)
I don't think you understood my post. I'm saying make a quality autonomous motion system independent of our modern roads, and then build a system to allow vehicles to deal with modern traffic rules and realities on top of that so when the present condition differs from what we have today, it doesn't become a matter of re-inventing the wheel.
I'm not saying throw out every bit of the present transportation system right away. I'm just saying that we don't need to entrench them as the entire basis of how the
Re: (Score:2)
Until we get flying vehicles we are going to have to deal with the constraints of roads. Bees and fish do not have those constraints and are a poor example to model road movement after. How do you build a highly constrained 2D system (roads) on top of a very open 3D system(flying or swimming)? The movement and constraints are completely different issues and the solutions are not transferable.
To paraphrase. Lets model autonomous systems now in an environment that may never exist in such a way as it will not
Re: (Score:1)
You're being quite short-sighted here, assuming that such a system would be restricted to cars. Why would we not expect the same system to be useful for helicopters and aircraft?
Re: (Score:2)
The article is about autonomous ground vehicles which is why I am discussing autonomous ground vehicles and how natural systems are not suited for autonomous ground vehicles. Had the article been about autonomous flying vehicles I may have had a different opinion.
Re: (Score:1)
I see no reason to let TFA negate the relevance of related technological advances and crossovers of technological investments.
Re: (Score:2)
Agreed. Take advantage of systems that have had hundreds of millions of years of evolution, rather than base it on human codes and rules that - intrinsically - are irrelevant to the "world" of autonomous vehicles.
Certainly, there will have to be some concession to human norms, as these vehicles will share the road with human-driven vehicles for a long, long time.
Nevertheless, count me as one of the people who feel that despite an almost-certainly-painful teething period, computer-controlled cars will be bo
Well, at the speed those robots are moving (Score:1)
You might be able get out of the way [youtube.com]
Problem is sensors and processing. (Score:4, Interesting)
They need a system that can identify visually, as well as with radar that can not be fouled by other radar to scan the road ahead. Because I can see scumbags setting up a radar broadcaster in their junker to cause an autoguide car to hit them for insurance money.
Re: (Score:2)
They need a system that can identify visually, as well as with radar that can not be fouled by other radar to scan the road ahead. Because I can see scumbags setting up a radar broadcaster in their junker to cause an autoguide car to hit them for insurance money.
Oh yeah ... this is a "hacker"'s dream.
"Oh, I was just trying to be helpful and find vulnerabilities, that's why I was remote controlling Grandma's car."
Re: (Score:2)
"How would that be different from causing someone to hit your car by turning the lights off on a dark road?"
My car has headlights and I can see the car as I approach it. Unles they painted the whole thing flat black, the reflectors built into cars will light it up 1000 feet away.
Now if you are a dolt and drive at 90mph into the dark with your low beams on... you deserve the accident.
Not so fast (Score:2)
This is all fine until someone gets hurt (Score:1)
Just because you can get a small number of autonomous robots to not crash into similar sized objects in a controlled environment does not mean it's ok for giant metal death traps to careen down streets and (usually) not hit small children darting into the street when their ball rolls away or as they bike.
It's all fine and dandy until someone gets hurt. And the second it happens to a little kid, it doesn't matter that you have 99.999 percent uptime, that 0.001 percent exclusion means obsessed parents will sh
Great Canadian X-Prize for autonomous vehicles. (Score:2)
All I have ever seen with autonomous cars is controlled tests in controlled areas under ideal weather conditions in Nevada or California.
It's time to put up or shut up about autonomous cars and put them under some real world testing.
Put 100 of these things on the Trans-Canada highway and ask them to drive from St John's, NL to Vancouver, BC in the middle of January and get there faster than a horse drawn carriage. Use the schmucks who want to fly to Mars as their passengers because they have no value for t
So many things wrong. (Score:2)
Fish and bees do not have to deal with two dimensional intersections.
When fish or bees touch each other it does not cause an accident that could kill someone.
Fish and bees are not constrained by the width of roads.
Vehicles move much faster than fish or bees and momentum is a much bigger issue.
Fish or bees generally do not have individual destinations which require movement in many different directions.
As for the demonstration, it was laughable.
Extremely slow.
No crossing.
No joining or leaving.
The term "zippi
Not WALL-E (Score:2)
Why is it that if you put even the dumbest electronics into a package that looks sleek and smart people think that it must be 'intelligent'? They did it a long time ago, stretching back to the mechanical turk I guess, where if the machine just looked human then it must contain advanced intelligence. Even these days whenever the Japanese put out some terrible uncanny-valley type robot that can only sit there and try not to be outwitted by SIRI people will still think it's some kind of advanced WALL-E style i
Build it and they will come (Score:1)
Re: (Score:1)
Go home Miley you're drunk again!