Driverless Cars

[3930dave]

@Level 5, it looks like the only thing that humans will be doing is specifying the destination.

Path to Autonomy: Self-Driving Car Levels to 5 Explained | Feature | Car and Driver

Arrested for slurring...... maybe ?

Rgds, D.

 

[rd_macgregor]

Yeah. I guess you'd have to "train" the voice recognition system to understand the slurred words (Hint: "home" is easier to say, drunk, than is "7354 Excelsior Lane")

 

[houstonscott]

No one wants them, not going to happen, what a waist of time.

 

[3930dave]

So if the operator needs to remain vigilant and be able to take over at an instant, what is the purpose of a "driverless" car in the first place?

If you think about it..... what the end-user really needs is the exact opposite...... expert systems that can take over when the human isn't able, for whatever reason.......

"Sucks to Be You" announcement from a computer doesn't help much, when the brown-stuff is about to hit the fan.....

Rgds, D.

 

[k0ua]

If you think about it..... what the end-user really needs is the exact opposite...... expert systems that can take over when the human isn't able, for whatever reason.......

"Sucks to Be You" announcement from a computer doesn't help much, when the brown-stuff is about to hit the fan.....

Rgds, D.

Exactly. The very thought of an expert system that drives during normal times and then bails on you JUST when things get rough is not liable to be successful. These designers may know a lot about robots but they know next to nothing about people.

 

[quicksandfarmer]

If you think about it..... what the end-user really needs is the exact opposite...... expert systems that can take over when the human isn't able, for whatever reason.......

"Sucks to Be You" announcement from a computer doesn't help much, when the brown-stuff is about to hit the fan.....

Rgds, D.

I think remote piloting is the answer. You know how a bunch of guys in trailers in a parking lot in Las Vegas are running our drone air war in the Middle East? Imagine a call center where all they do all day long is take over the piloting of driverless cars when the on-board controls can't figure out the situation. If it can be driven by a computer it can be driven over the network. Plus the guy in the call center would have the advantage of all of the onboard sensors.

Obviously that won't work for crash avoidance. Thinking of the fatalities so far, the system needs to be more cautious, more inclined to switch to manual when it encounters a situation it doesn't understand. Imagine a system where it's on automatic 95% of the time. One guy in the call center could cover an average of 20 vehicles. That's still a tremendous labor savings over one driver per vehicle. The trickiest 5% of driving is going to include a lot of routine stuff for a human, things like left turns across traffic. The safest 95% is pretty achievable for a computer.

 

[Dadnatron]

Big Equipment, Powerful Diesels, What Could Go Wrong.....

....... Go Wrong.........

Wide open fields.... not much to hit. Gated community, relatively low speed and complexity. Not bad places to start.

Rgds, D.

The question won't be are the 100% safe... it will be are they safer than a human. Humans are notoriously fickle. It won't take much to make a machine which can top our concentration.

The real trick will be getting a 'critical mass' of interconnected cars. When there are enough cars all talking to one another, letting each other know what they are doing, most issues will subsequently resolve. I foresee a specific 'town/county/area' going totally automated, which will remove the human error. When it works, and I believe it will work well, it will become more and more widely adopted.

I spend a lot of time behind the wheel, and for me, I welcome total automation. I foresee me hopping in the pod-mobile, punching in the destination, being whisked away into a 'stream' of other automated cars so close as to be drafting off each other. All are talking to each other, so all cars 'behind' knows what the leaders are seeing. "object to the left of road" "Pothole 3' from left edge" "debris in road" etc. every car within 1/2m is talking to one another and so nothing is a surprise. At THAT POINT, true automated travel can occur.

But the test will not and should not be perfection... it should be 'better than human'.

 

[3930dave]

I think remote piloting is the answer. You know how a bunch of guys in trailers in a parking lot in Las Vegas are running our drone air war in the Middle East? Imagine a call center where all they do all day long is take over the piloting of driverless cars when the on-board controls can't figure out the situation. If it can be driven by a computer it can be driven over the network. Plus the guy in the call center would have the advantage of all of the onboard sensors.

Obviously that won't work for crash avoidance. Thinking of the fatalities so far, the system needs to be more cautious, more inclined to switch to manual when it encounters a situation it doesn't understand. Imagine a system where it's on automatic 95% of the time. One guy in the call center could cover an average of 20 vehicles. That's still a tremendous labor savings over one driver per vehicle. The trickiest 5% of driving is going to include a lot of routine stuff for a human, things like left turns across traffic. The safest 95% is pretty achievable for a computer.

Interesting. ONStar on Steroids.

The technology is easy, not sure the rest of the recipe will line up...... Military is motivated to use the technology largely for cost reasons - pilots are expensive to train, expensive to replace, and there are political costs associated with loss of life in combat.

Civilians on roads..... well, bluntly put, we're cheap to replace. On a personal level, I don't agree with that last sentence but I do recognize the reality.

Rgds, D.

 

[3930dave]

But the test will not and should not be perfection... it should be 'better than human'.

Realizing how low the bar is actually set, I decided to pay more attention to what is emerging in this field.... it's probably going to change faster than we think.

Jump forward even 25 years, few people will pay much attention to fully automated transport - it will just Be There. How many people today think about how their fuel injectors in their cars work, or how GPS functions ? <- those are There Now.

The interim transition will be bumpy - tech issues need to be sorted out and well proven, but probably the bigger and longer battles will be fought in court - Who is Liable, and most importantly, Who Pays ?

The insurance industry became wealthy taking small amounts of money from large groups of people - this model may end up disrupted and/or broken. Even govts have a tough time getting money out of large corporations......

Rgds, D.

 

[Dadnatron]

If you think about it..... what the end-user really needs is the exact opposite...... expert systems that can take over when the human isn't able, for whatever reason.......

"Sucks to Be You" announcement from a computer doesn't help much, when the brown-stuff is about to hit the fan.....

Rgds, D.

This is already occurring with LIDAR assisted automatic braking. It will happen, but there needs to be a standard onto which ALL vehicles are based.