Limitations of "radar cruise" augmented driving

Just old-school cruise control, no radar, maintains a set speed, that’s all.

 

I believe that when the programmers of the machines get to this, they will find that humans are not as good at divining the situations and intentions of other humans as they think they are. Humans make too many errors, both on the divining side, and on the "another human" side in following through on the clues they emitted. And other surprises, around the corner and out of sight, will force some unexpected stops or slowdowns.

For overall safety reasons, the machine responses will be to follow proper defensive driving practice by preparing for multiple possibilities, including the low but very real likelihood that the car ahead will not clear the lane when originally expected, or at all.

This practice will leave numerous drivers frustrated with "unnecessary" slowing by the "dumb" computer on the many times the lead car does clear in time, but will reduce the rate of more drastic responses, and crashes, when things don't go as hoped.

Human drivers in the U.S. are currently killing over 100 people per day. If done correctly, computerized "self driving" cars should be sharply reducing this number.

 

Key word...
"should"

To be fair, I think that most of the OEMs are doing what they can, but you just can't fix stupid.
There are more cars on the road, and there is an undeclared "arms race" in BHP, and for lots of reasons road rage seems to be more and more of a thing.
Oh....and in the Land of the Free all you need is about 16 candles on your birthday cake and about a 6th grade reading ability to get a driving license - and in many states there is NO maximum age limit.
AND.....cars are getting both lighter AND larger.

I think that this is a pretty dangerous environment to be giving people the notion that cars can be "self driving"

Safety features CAN make cars safer to drive.
Whether or not this is actually occurring in the real world is something that I'll let statisticians fight over...

 

Yes. If you hold the button in for a few seconds it activates "dumb" cruise control.

 

This implies that dumb humans aren't programming the computers or making the components. A self-driving car can just as easily suddenly perform a maneuver that was unexpected to other self-driving cars as a human. It's also just as easy for a self-driving car to fail to divine the intentions of the other self-driving cars surrounding it. Yes computers can react more quickly, but they can only react in ways they are programmed to react, and there is only so much they can do in a given situation given the physical constraints imposed by momentum.

 

Yes, when they are perfected self driving cars will be safer AND more efficient. Perhaps there will come a day when only suitably equipped cars will be permitted on certain roads. It will be like having a chauffeur, get in and get out, won't that be great?

For now efficiency vs safety is a choice engineers and each individual driver must make, and choosing a one size fits all answer will be difficult. The inputs to that equation are infinite. Are you out for a relaxing drive or rushing a bleeding child to the hospital? I think that for the foreseeable future humans, although perhaps not as good as they think they are, will still be better than algorithms. Of course "the low but very real likelihood that the car ahead will not clear the lane when originally expected, or at all." is there, and if that low probability event happens I will have to brake harder than I like, or change lanes, but I will not have to drive into another vehicle or into the ditch. I take all that into account, mostly subconsciously and still drive safely.

My definition of safe driving includes the axiom that "No other driver should have to change their planned trajectory due to my actions". Some people get very upset and honk their horn if someone quickly pulls across their lane to hit a gap in traffic. I don't, I consider it perfectly fine, efficient even, as long as I don't have to brake to avoid them. If I do have to brake, even a little, to avoid a collision, then is was reckless and stupid. I think you get my drift.

I love automatic braking, I think it is a wonderful safety feature that will save lives. If a pedestrian or bicycle (TSS 2.0) appears in front of me the car can react faster than I can. If I am just driving down the road and I see a situation where my human intelligence tells me there is no danger in NOT braking at this moment, I will happily override it.

 

In my experience support for mandatory self-driving cars is inversely proportional to one's understanding of computers and programming in general. Most people do not realize the monumental difficulty of designing a system with the flexibility and capabilities of even the "dumbest" human driver.

Look how "stupid" our Primes are when all they're trying to do is stay in a lane and nothing else, or the DRCC trying not to hit the car in front of you. Can you imagine adding a mountain of complexity on top of that?

 

Dumb mode is when it does not use the radar to synchronize your speed with the car you are following. Hold down the button on the end for 5 seconds as you turn cruise on and you enter "normal" cruise control mode.

I have the same problem. My old car was a 2002 gen 1. My wife's is a 2010 Camry hybrid. It's disconcerting to have the same control do different things on the two cars.

 

No it isn't. You push and HOLD the set button (speed down) until you reach the desired speed......and then release it.

 

Well then it's no "smarter" than human drivers.......who tend to follow the car in front of them right off into the ditch during times of low visibility.

 

I tend to disagree. It's not a matter of programming so much as it is a matter of system design. Most efforts are limited to making the car behave like a human. As a result many use visual cues to decide what to do, because that's what people do. It's a lone agent in a chaotic environment.

What will make the major change is when someone gets the authority to design and implement a system where all cars (including those driven by humans) have transponders that allow cars to communicate. When 3 cars approach a blind intersection, it would work so much better than humans if the cars will simply negotiate which ones will enter the intersection first. Combine that with other capabilities and things get interesting. Imagine what you get when instead of brake lights, each car also sends a data packet that identifies which car it is and what it's about to do? Instead of a 5 car pile up when a puppy runs across a busy street you can have all 5 cars stopping at the same time and at the appropriate rate of speed. No rear-enders. No dead puppy.

A lot of the uncertainty of self driving cars is due to humans driving badly and unpredictably. Imagine how nice it will be when every car that approaches the left turn actually signals to the surrounding cars? And how nice it will be when every car with a flashing turn signal actually makes that left turn?

If you think about it, the current self driving efforts do a great job of driving as well as a perfectly legal bad driver. Some do better than an average driver. But to make people happy they need to do much better than a very good driver.

 

As an add on.... I forgot to explain that I was once trained to do systems design and systems integration. It's drastically different from just programing. If you design a system, you can get rid of a lot of the chaos, even to the point where the biggest obstacle becomes mechanical failure. Well designed systems handle tasks much better than their human counterparts.

It's much more problematic when you try to program a device that has to interact with the environment while the environment has no restrictions as to what it does to your device.

 

Did you read the entire thread? The question was not how to rapidly change the set point, the thread was about how much faster the set point changes than the car typically decelerates. One answer was to decrement the set point one mph at a time so as not to outpace the actual speed decrease. As was pointed out, doing this from 65 mph to 40 mph requires 24 down presses. If you simply push and hold the speed down you will reach a set point of 40 while the car typically is still doing 60 which will result in sudden braking when you release the set button.

 

1. That's pie in the sky stuff, though (and I'm trying to avoid the political implications of a subset of the population that truly believes you can nullify human nature with laws and regulations). That's like saying "imagine a world where no one's mean to each other." It's easy to imagine, but it's never going to happen. Also implicit in that world of flawless self-driving cars is the requirement that the human occupants have no ability to change course or override the car's behavior. After all, if the humans are allowed to change their mind about where they're going it will return the predictability problem back into the equation.
2. The transponder idea only solves part of the self-driving car problem. It doesn't solve the problem of unexpected obstacles. If a tree branch falls onto the road does the car swerve into oncoming traffic, into the pedestrians on the sidewalk. or does it kamikaze the driver into the branch? How will the system understand context, such as a soccer ball suddenly rolling out into the street. How will the system know that means a kid is about to follow? What about items falling off the back of someone's pickup on the highway? What about malicious vandalism of road markings or signs (especially with cars that human occupants can't override). There's just countless road dangers that technology is nowhere even close to being able to address effectively.

If I have a car that won't let me decide to turn left when the pre-programmed route wanted to go right, or is going to decide for me that it's more important that I die as opposed to taking some other evasive action I'm going to burn that car to the ground, hopefully coasting it through the front doors of the dealership that sold it to me while it's burning. I'm not alone.

 

A great post on next-level system thinking, but the plot thickens rapidly. Perhaps should substitute "force" for "allow" as it doesn't work nearly as well if car #1 is not participating. Also let's not forget communications security. We will have spoof-proof encrypted and authenticated packets I assume, and jam proof as well? Big prize to the first hacker that can command all cars to move out of their way so they can go 100 down a busy 8 lane freeway.

 

You've never been in a train or a plane? In both cases you have no direct control over what the driver is doing.

You might not realize it, but your driving today is being controlled by computers. The traffic lights are often networked and direct traffic to accomplish the objectives that the city has set forth. In some cities, the traffic signals are timed to encourage people to use the freeway. In others they are timed to get you through with minimum stops to minimize pollution.

It's not hard to network the cars and have them work cooperatively. It's actually quite easy and has been done in simulations many, many times. It's even done in video games like Grand Theft Auto, although with different intent.

The question of the fallen branch is an interesting one. Today, when that happens people do whatever they want to. They hit the branch. They drive off the road. They drive on the sidewalk. They panic stop and get rear ended by the person behind them who can't see the branch. Or, if they are lucky, they slow down and stop and no one hits them, so they turn around and leave the obstacle for the next driver to deal with. With networked cars the first car to encounter an obstruction can and should broadcast the need for a detour so all other cars would not hit it. It would be nice if the highway department was monitoring those transmissions so that the tree could be removed ASAP.

I've designed and programed systems that worked very well in a chaotic environment. It's amazing how well you can make it work as long as there is a valid action that can be taken when the SHTF. After all, it does not matter what falls from the sky, whether it's a tree, a plane or a bus since you'll take similar actions.

Some people hate the idea of autonomous cars, and some can see the value. The difference is often a perception that the drivers lose an essential liberty to control the gas pedal. Personally, I don't have that fear, so it's not an issue.

 

Yes but my comment was addressing only that one isolated statement.......not the entire subject.
And some systems reduce the set speed in the blink of an eye.....and some don't.
But you are right in that doing it that way likely will result in a rapid deceleration which you don't really want.

 

I do not own a train or airplane. If I did I would expect to exercise control over them. To answer the question, no I do not ride either very often and mainly for that reason. I haven't taken a flight in over 20 years. I do ride the train on occasion, but only for scenic trips. I do not ride commuter trains.

As with everything these days, there's even a political agenda when it comes to traffic signals. In Los Angeles they're intentionally sabotaging the smooth flow of traffic in any way they can to create demand for massive condominium projects their real estate developer cronies build. In return they get campaign contributions, donations to their non-profit organizations that employ their useless friends and family, and secure private sector jobs for themselves after they term out.

They remove traffic lanes from arterial roads, obstruct the flow of traffic by removing left turn lights, set traffic light timing to increase gridlock, refuse to maintain streets ... the list goes on. Of course, public transportation ridership still continues to decline, so all they've really managed to accomplish is to reduce everyone's quality of life and increase emissions. Doesn't matter to them as long as those campaign dollars and lucrative jobs for friends and family at non-profits keep flowing. That's precisely the reason those people should be prevented from having further control - because inevitably they will abuse those powers for their own benefit. It's human nature and cannot be stopped save for denying them access to power in the first place.

A perfect example in California comes from something as mundane as a thermostat. No sooner had WiFi thermostats hit the market than California politicians began attempting to pass legislation allowing them to legally hijack your thermostat for the "public good." If they determined the temperature you set was not to their liking they would be able to take control of your thermostat in your home and set it however they feel is more appropriate. That's the number one reason I refuse to allow any appliance or controller that has internet access into my house.

I do not "fear" the abuse that will inevitably follow if private means of transportation even has the capability of being controlled remotely. It's a virtual certainty, and one that I will not allow.

 

And pedestrians; when you’re trying to cross an intersection and the oncoming drivers are too brain-dead lazy to signal which direction they’re going…