Honda Crash Detection System

MImeKillEr writes "MSNBC is reporting that Honda Motor Co. unveiled an early crash-detection system for one of their vehicles. The system is unique in working even before the driver responds. A radar in the front of the car stashed behind the Honda logo detects vehicles within a range of about 300 feet ahead. It then taps the brake and tightens the seatbelt. A buzzer goes off and a light on the dash is illuminated. If the driver responds, the braking power is boosted. If the driver fails to respond, the system kicks in and brakes more while also tightening the seat belt. Unfortunately, Japanese regulations don't allow for the system to fully stop the vehicle."

Comments

DOes it work ?

[dargaud]

Now I write radar software and I really don't have a clue how such a system can work reliably. A non-moving car 100m ahead ? That happens every time there's a parked car in a curb. Car radars are not like aircraft radars. The latter only has to see something ('anything') in the middle of a big mass of air. Nothing else around. A car radar would have to sort out lots of echos at various doppler: the ground is coming towards you (when it's far ahead), other cars going the same direction (slower in the right lane, faster in the left lane), cars coming the other way, parked cars, things hanging overhead (bridge, street lights. advertisement...)

Just imagine driving on a mountain road and out of a right curb comes a car driving the other way. The radar sees it right in front of you, coming your way. How does it react ? I'd hate to see it break suddenly, particularly if the road is wet or snowy.

Re:DOes it work ?

[Brento]

Just imagine driving on a mountain road and out of a right curb comes a car driving the other way. The radar sees it right in front of you, coming your way. How does it react ?

It works fine. Check out the radar-based cruise-control from Mercedes, now available on a few models. You can set your cruise to follow a vehicle ahead automatically. You just steer, and the two (or more) of you can pass cars and go through tunnels just fine without the cruise control panicking.

Re:DOes it work ?

[Lumpy]

Now I write radar software and I really don't have a clue how such a system can work reliably.

Oh come on... either you are lying to us and you just obviousally told all of us that by making that statement or you dont have enough coffee to think this morning...

There are 2 ways of doing this.

1 - simply measure the speed of the car in front of you relative to you. if the car is doing 50mph toward the radar source, hit the breaks! you simply have a threshhold setting for distance to trigger and speed. and simply adjust that according to the car's current speedometer reading.

2 - to hell with the speed of X Y or Z look at the doppler shift. if I detect a positive doppler shift of a magnitude above the setpoint... trigger ..

this way no matter what the distance it will trigger if it senses the target is closing fast to react.

It wont stop you from bumping the guy in front of you at 10mph... but it will stop you from flying into your parking space at 70.

Re:DOes it work ?

[valkraider]

Many people are misinformed about the newest radar systems. With new technologies, police can do a lot more with "regular" radar than most people think. Check out Radartest.com for more in depth info and some good links.

Re:DOes it work ?

[Titusdot Groan]

How did this get an informative? This is an obvious (deliberate?) misreading of the parent post!

The work reliably part had nothing to do with speed it had to do with orientation -- how do you tell the parked car 100m ahead is in YOUR lane -- radar can't read the lines on the road and tell that your lane is about to curve and take you safely BY the parked car.

Re:DOes it work ?

[PD]

Not a problem, do what police radars already do. Don't measure the doppler shift, instead combine the waves of light with light of a fixed and known frequency. Then, you measure the beat frequency.

Re:DOes it work ?

[alienw]

How the hell do you think radar works? It usually measures doppler shift of a reflected radio signal. That's how, say, police speed guns work, which do basically this exact thing. "Lidar" guns (laser) simply bounce a laser beam off a car a few hundred times a second and count the number of nanoseconds it takes for it to bounce back. It's millions of times easier than measuring doppler shift of light, anyway.

Re:DOes it work ?

[WolfWithoutAClause]

Wrong. This is a huge deal if you happen to have seen the problem 125m ahead and are taking evasive action other than braking, for example changing lanes or even swerving. The slightest braking during a sharp direction change can throw your car into a skid.

Ok. Good points. But I have a few observations:

a) if the vehicle has ABS (I assume it has), then a skid is unlikely

b) The system should be able to detect the cornering and adapt to it anyway to a large degree

c) in most cases a modest reduction in speed will greatly improve cornering anyway- maximum cornering goes as the square root of your speed- a reduction of 10% gives ~5% tighter cornering whilst reducing the energy of the car by ~20%.

Re:DOes it work ?

[Jerf]

Oh come on... either you are lying to us and you just obviousally told all of us that by making that statement or you dont have enough coffee to think this morning...

No, the problem is you obviously have no experience dealing with real world data. Writing software is a lot harder then "seeing a way to do it". For instance...

1 - simply measure the speed of the car in front of you relative to you.

Not even all the way through your first sentence and you've already hit major problems. First, all measurements have noise in them; you can't know the "speed" with perfect reliability, and how you determine the speed from the incoming data is a non-trivial problem. Since lives are on the line, it's important to get it right. A simple "moving average", for instance, won't work because it will slow the reaction time of the system down unacceptably.

Also, we have a problem with "car in front of you". Now, "thing in front of you" may work, but A: how do you know it's a car? and B: how do you know it's truly in front of you? Maybe you're just turning and that "thing" you see coming at you is just a mailbox.

Pop quiz: You are a radar system. You detect something coming at you at 70 mph from a hundred feet away. You're turning slightly, but not much. As a radar system in a car, this is all you know. Do you slam on the brakes?

Yes? Oops. Turns out that you're on the left lane of a three-lane highway, and somebody's parked their car on the right shoulder due to car problems. The car gets rear-ended by a tailgaiting Mack truck and the driver dies. You get sued because your system was not supposed to hit the brakes then.

No? Oops. There was a Mack truck jackknifed across the road which the driver wasn't looking at and didn't see. In the .4 seconds it takes for the driver to react that you could have been braking, the car continues at 70 mph. By the time the driver reacts, its too late to slow the car down to a survivable speed. The driver hits the cab of the truck and dies. You get sued because your system was supposed to hit the brakes for the driver.

This isn't so easy, is it? The car only has a very limited number of senses; in human terms it's blind except for one whole pixel, deaf, and can only feel limited information about it's own state ("I'm going 50", "I'm turning", etc.). It's not like the car has a picture of the whole road to work with... and of course the reason it doesn't have that is we wouldn't know what to do with it if we did. Computer vision is still a very limited technology; it can not handle the real world to any significant degree yet.

if the car is doing 50mph toward the radar source, hit the breaks!

How hard? What threshold for distance? Does the distance threshold change if the relative velocity changes? How does the brake hitting relate to the relative speed? What if I'm driving down the road at 60mph, I go into a turn, and a mailbox momentarily looks like it's coming at me at 55mph? Do I slam on the brakes then, or wait until I'm sure it's a car? How can I be sure when all I have is a distance and a relative velocity anyhow?

If you're wrong and you "hit the brakes", you may end up killing the user. Lives are on the line; you can't gloss over any of these questions. One wrong answer and people die.

you simply have a threshhold setting for distance to trigger and speed.

That is naive beyond words. The real world is much more complicated then that.

and simply adjust that according to the car's current speedometer reading.

How, exactly? "adjust" is an empty word until you give more specification. It's damned easy to say "simply adjust", but there are hundreds of plausible way to "simply adjust", and most of them are wrong. (It's even possible all of them are wrong; there's no guarentee this system is even possible, you know.)

2 - to hell with the speed of X Y or Z look at the doppler shift. if I detec

Re:DOes it work ?

[JerkBoB]

Aren't you supposed to turn your car off for those carwashes? It's not like the wipers are going to run when your car is off.

Nah, at every one I've been to the monkey running it has just asked me to put it in Neutral and keep my foot off of the brake. How else would I listen to the cd player? :)

NOTE TO /. COMMUNITIY: THIS IS HOW ONE SPELLS THE WORD FOR THE PEDAL NEXT TO THE ACCELERATOR!!! There has never been a 'Break' pedal in any car I've driven. Sorry, I've just seen that too many times in the comments for this story.

Mercedes System already does this

[servicepack158]

The AMG benz's come with a cruise control system that when active will slow down if a car in front of you is going slower then you, and speed back up to cruise speed when they get out of your way. It can apply up to 20% of the breaking force of the vehicle too if there is a sudden stop. you can look at it and watch a flash presentation on the website. To the dork programmer who doesn't understand how it works: That's why honda didn't hire you for the project :) You only activate such a system above certain speeds. So basically it's an autopilot system for a car, but people always change lanes without looking so now they need to invent something to lock the steeringwheel :D

Re:Whatever happened to Volvos being the safest?

[AlinuxNCSU]

If you recall, Volvo was purchased by Ford. I don't have anything against Ford (my family has owned three Tauruses), but they don't have the same reputation that Volvo had. It was at that point that Volvo's styling began to change, going from a less boxy style to the more consumer-friendly styles of today.

Since then, Volvo has shut down it's Swedish offices and moved headquarters to California. They have always had the reputation as safe and conservative in design, but Volvo is not the same company as before -- they are a Ford-owned American car company that uses the Volvo name. I won't pass complete judgement, not having followed Volvo's production for the past couple years, but I would venture to guess these basic facts have something to do with your complaint.

-ALinux

Here's a link to more info from Honda

[zptdooda]

Hereâ(TM)s more from Honda:

CMS

So itâ(TM)s more than just the 300 ft test, which would be arbitrary. It looks at "distance, speed and and anticipated path".

Sounds worse than a backseat driver though.

Re:Being cut up

[AlinuxNCSU]

For those of us not familiar with the idiom you're using, we would use the term "cut off".

Re:300 ft ?

[turgid]

It depends how fast you are going. 300ft at 70mph is approximately the minumum safe distance to leave between cars to enable them to take the appropriate action in the event of an emergency. If everyone knew this and stuck to it there would ba a heck of a lot less accidents on the motorways.

Re:300 feet?

[dex22]

You don't like seatbelts, and you claim to have exceptional response skills.

It is for people like you, who think they're great drivers so don't pay attention, that this system was developed.

Re:I can picture this...

[Jens_UK]

Four point seat belts wouldn't castrate you; you're thinking of five point belts.

Re:Instead of braking...

[SnowDog_2112]

It starts with an alert chime. It doesn't apply the tactile feedback (seatbelt+brake) unless you fail to respond and it feels a collision is still likely.

Honda has a whole page about this feature; check Google (or just read other posts in this thread, it's been linked twice already that I've seen).

Re:ABS Works

[Alan Partridge]

"Oh, how did you figure that braking and releasing would ever be better then constant braking. I mean if the wheel isnâ(TM)t moving wonâ(TM)t it have more friction then if the wheel is allowed to rotate, even a little?"

Errmm... the REAL point of ABS is to allow you to steer - and thus retain control of - your vehicle under emergency braking. It's saved my life TWICE.

I remember this project at Penn State

[Anonymous Coward]

Dr. Tony Ferraro worked on this project along with my current advisor, they actually used a radar device placed in a round container stuck to the front of a car...crap was bigger back then... from the videos I saw it worked quite well. It actually compared the present situation with an enormous database full of other situations. It did set off the alarm when they quickly approached a guard rail on a sharp turn though. Among other things, it knew current speed and acceleration relative to objects in front of the car.

Re:Control over the vehicle

[NukeIear]

Of course ABS was NOT designed to make you stop faster. Though in theory it could make you stop faster, something you might have proved in highschool physics. But the world is a bit more complex then HS physics problems, so sometimes abs is faster, sometimes standard brakes are, depending on road conditions and who is driving.

Big non-boonies highway patrols still use RADAR

[Chazman]

CHP (California Highway Patrol) uses Ka-band RADAR almost exclusively. In fact, in the last five years of driving around southern California, the vast majority of the detector hits I've gotten that I could pinpoint to a particular law enforcement source have been Ka-band RADAR. I've only been hit with LIDAR once. That was a Newport Beach city police motorcycle unit. How do I know this? I've got a Valentine-1 and a Lidatek Laser Echo, and I keep them both on nearly all the time.

Yes, LIDAR has greater range and greater selectivity (can pick out a single vehicle) than RADAR. But it also requires a stable platform and sighting equipment to be used properly. It cannot be used from a moving vehicle. RADAR can. It cannot be used in a shoot-from-the-hip quick reaction scenario. RADAR can. It cannot be used without a sight attached to a stable semi-fixed platform. RADAR can. For these reasons, the demise of RADAR is vastly exaggerated.

Big trucks have had this for a while

[DonGar]

They've had this for a while on the big trucks, where the system costs in the $100k range.

They really seem to work, because they've been shown to boost the safety records of the drivers/companies that use them.

I used to have an article about it laying around. I dug it up when a friend and I were discussing what it would take to really build a self-driving vehicle for mass production.