On robocars, congestion and road capacity

2 vues
Accéder directement au premier message non lu

Brad Templeton

non lue,
15 nov. 2010, 20:10:5615/11/2010
à transport-innovators
I've written a few times in this newsgroup about the potential for
robocars (and ITS) to significantly increase the capacity of the
existing roads, also greatly reducing traffic congestion. This
turns out to be fairly important in contrasting robocars (which offer
PRT on existing streets) with elevated guideway PRT (which in many
cases uses automated rubber tire vehicles.)

There are a wide variety of factors which can increase road capacity
on the existing infrastructure. Together I believe this could result
in roads being able to move 5 times as many people as they move today,
and in theory even more. (In fact, in the extreme theory, over short
high need stretches like single arteries, the use of ad-hoc buses
could move 1 million people per hour through a 6 lane -- normally 3
each way but redirected at rush hour -- artery, compared with about
9,000 per hour for a 3 lane each way artery.)

However, without need to get into such extremes, it seems the capacity
to meet current and future needs for some time is already present
through a wide variety of techniques. Some can start improving things
very early, others require a road with mostly robots and few humans.

This is all discussed in this essay:

http://robocars.com/congestion.html

And there is an associated blog post for comments:

http://ideas.4brad.com/robocar-impact-traffic-congestion-and-capacity

I'm certain based on past traffic here that these plans will be
controversial, so let me add one note that I do include at the end of
the essay. It is possible, if it is needed, to build inexpensive
guideways for use only by narrow, lightweight single person robocars
in the most congested parts of cities. These guideways would cost no
more than typical PRT guideways -- possibly less than some as they
would be simple and unpowered -- but need only be built in a limited
subset of places, while the robotaxi network is able to take people
from any point to any other point directly and with minimal
congestion. The guideways could also carry other vehicles, such as
delivery robots and even pedestrians and cyclists during off-peak
hours when the cars can remain on the streets.

This is a powerful "best of both worlds" approach, if we need the
guideways at all. They certainly aren't needed in the residential,
industrial and suburban zones.

Dennis Manning

non lue,
15 nov. 2010, 22:30:5015/11/2010
à transport-...@googlegroups.com
The most widely accepted definition of PRT states that PRT must be separated
from existing street traffic. Robocars on existing streets is not a PRT
system. Let's not confuse what PRT is.

I don't see substantial congestion relief until the day that robocars can
escape the limitations imposed by mixing with manually operated cars. That's
a long way off.

--------------------------------------------------
From: "Brad Templeton" <bra...@gmail.com>
Sent: Monday, November 15, 2010 5:10 PM
To: "transport-innovators" <transport-...@googlegroups.com>
Subject: [t-i] On robocars, congestion and road capacity

> --
> You received this message because you are subscribed to the Google Groups
> "transport-innovators" group.
> To post to this group, send email to
> transport-...@googlegroups.com.
> To unsubscribe from this group, send email to
> transport-innova...@googlegroups.com.
> For more options, visit this group at
> http://groups.google.com/group/transport-innovators?hl=en.
>
>

Brad Templeton

non lue,
15 nov. 2010, 23:13:4015/11/2010
à transport-innovators
Well, the essay details a lot of ways it can happen while there are
still many human cars on the road. Some of it comes from turning
the humans into "robots" when it comes to choosing their route, which
is to say they obey the nav computer not only when it tells them what
streets to take, but when it tells them they must wait before
departing the house (ie. metering lights, but in your car, not on the
on-ramp.)

Proper metering is the key to congestion, which is caused by having
more cars on the road than the road can handle, combined with
"incidents" which drop that capacity number for various reasons.
Some of this is not all or nothing. The more robots, the fewer
accidents, and that's one of the biggest congestion causing
incidents. The more robots the less irrational slowdown. You don't
eliminate it, but you reduce it. It's not like it's a step function
waiting for robots to go from 99.9% to 100%.

More use of half-width cars naturally increases capacity, and again it
is no step-function. Robots can follow human cars at a shorter
headway because they have the reaction time for it, while humans must
keep a human reaction time headway. Again, capacity increases
gradually.

Dynamic street reversal, which can result in a 1.5x increase in
capacity at rush hour, is already done on some highways and bridges,

In fact, I would be curious as to which things you thought were the
step functions that didn't provide benefit until we had almost no, or
no human driven cars?

As to what is PRT, I will point to two things. First, the name
"Personal rapid transit" which does not include the words "grade
separated" in it anywhere.

Secondly, I will say that I believe that even if you have a definition
of PRT, what matters far more than that definition is the *goal*, not
the implementation. I believe the goal is the source of any valid
definition and that goal is:

a) Personal transportation -- you and your associates ride without
others.
b) A rapid trip, which means that it goes point to point without
stopping at stations, which also means not waiting for the vehicle,
not stopping at stations and yes, not being slowed much by congestion.
c) Available to the public (that's the transit part) walkable from
your home or other destination.

Guideways are one path to the goal. Well regulated roads can be
another. What really matters is do the riders think they got a nice
rapid trip.

The PRT vision has often been expressed even more succinctly as
combining positive values of public transit and private cars. I
think anything that does that is a form of PRT, and that it is wrong
to try to say that if it doesn't do it the way you imagined, it's not
PRT.

However, rather than arguing definitions, which gets nowhere, I am
more interested in the question I cite above -- which benefits do you
believe don't appear until the street is all robots, or mostly
robots? It's easy to demonstrate that a totally robot street can
have vastly more capacity and smooth flow than any guideway system
I've ever seen proposed, but that's uninteresting. It's more
interesting to ask what you can do and can't do at 10% robots, at 25%,
at 50%. The main thing that I can think of that needs 100% robots
would be a high-volume at-grade intersection with no stopping. Low
volume at-grade intersections can work fine in mixed traffic because
the humans still stop all the time, while the robots only stop if
there are humans present who have just stopped.

On Nov 15, 7:30 pm, "Dennis Manning" <john.manni...@comcast.net>
wrote:
> The most widely accepted definition of PRT states that PRT must be separated
> from existing street traffic. Robocars on existing streets is not a PRT
> system. Let's not confuse what PRT is.
>
> I don't see substantial congestion relief until the day that robocars can
> escape the limitations imposed by mixing with manually operated cars. That's
> a long way off.
>
> --------------------------------------------------
> From: "Brad  Templeton" <brad...@gmail.com>

Dennis Manning

non lue,
15 nov. 2010, 23:36:4415/11/2010
à transport-...@googlegroups.com
Speaking of definitions when you ask what can be done with 10% robots what
do you mean by "robot"? Is it 100% driverless? or driverless but with a
human that can override the computer? or?

If it's 100% driverless than I still say it's a very long way off.

--------------------------------------------------
From: "Brad Templeton" <bra...@gmail.com>
Sent: Monday, November 15, 2010 8:13 PM
To: "transport-innovators" <transport-...@googlegroups.com>
Subject: [t-i] Re: On robocars, congestion and road capacity

Brad Templeton

non lue,
16 nov. 2010, 00:25:4016/11/2010
à transport-innovators
There isn't a simple answer to that. Generally a robocar operates
without human intervention, and in fact it's fairly important that the
vehicles be able to move themselves while vacant. However, the
reason the answer isn't quite as simple as that is that on the path to
being able to do that, there are intermediary steps.

For example, if there is a human in the car, you can call upon that
human for decisions in unusual circumstances. Nothing real time, the
vehicle has to be able to handle any short-term tactical problems like
avoiding obstacles and getting to a safe state if a situation it
doesn't understand appears. However, you can certainly imagine
asking the occupant what to do if the car is to the first to discover
something new, like a policeman directing traffic or a barricade on
the street or an accident in the middle of the road. The car will
know to stop for these things, but might ask a human what to do about
them.

The human need not be in the car if there is network connectivity --
and we're moving to a world where there are certainly no urban streets
without network connectivity. The human might be remote and might
look at the video to figure out what to do. And this would in fact
be one solution for empty cars that are moving themselves for storage,
recharging or to pick up passengers. Empty cars, fortunately, do not
need to be able to go nearly so quickly -- they are not in a hurry --
nor do they have to go down every possible street except for the last
mile to pick somebody up.

The car that needs advice from its occupant rarely is not a long way
off, it's very close to already here. How often it asks for advice
affects how pleasant the ride is, but it need not be often, for once
one car has asked for advice about an unusual situation the next car
may not need to. The eventual goal, which is a few years down the
line, is for the occupant to be able to sleep, meaning you can wait 15
seconds to get their advice. (Again, only very rarely. If you have a
situation that you know needs advice, like a police officer directing
traffic, you can awaken the sleeper before you get to where you now
know the policeman to be based on the report of the last car. Only
the first car to see the cop has to pull over, pause and wait for the
sleepy head to get up.)

The goal of course is to reduce the number of situations that require
intervention over time until it gets very low and you can usually
sleep away the whole trip. (With remote video camera assist, that can
happen fairly early, because you only have to wake up if a hard-to-
figure situation occurs off network.)

I have another essay in progress, as it turns out, on the intermediate
world of human guided vehicles, which can range from a very early
"joystick driven car" to full autonomous operation at speed, almost
never disturbing the passengers.

I think we could build the joystick car right now with some basic
R&D. That actually could be highly useful to PRT. Imagine something
like Heathrow PRT where the pods can leave the track when they get to
the car park, instead of having stations, and be joystick guided to
the occupant's actual parking space. Joystick guiding means like a
video game. Point it where you want to go, and it will go there. It
will resist you trying to steer it where it might hit anything. And
it can slowly go back to good waiting spots or even watch for people
who put up their hands to ask a car to come to where they just parked,
or even notice people who are parking and will want a pod. That's all
something we could build today. It's more what people would want,
especially if they have luggage, and doesn't need stations so it can
save money.

As you know, factory floor and delivery robots already interact safety
with pedestrians, in part by going slow enough to be safe. The only
question is how fast can you go as the tech improves, until it's fast
enough to be useful for vacant vehicle moving.

On Nov 15, 8:36 pm, "Dennis Manning" <john.manni...@comcast.net>
wrote:
> Speaking of definitions when you ask what can be done with 10% robots what
> do you mean by "robot"? Is it 100% driverless?  or driverless but with a
> human that can override the computer? or?
>
> If it's 100% driverless than I still say it's a very long way off.
>
> --------------------------------------------------

Jack Slade

non lue,
16 nov. 2010, 01:40:1116/11/2010
à transport-...@googlegroups.com
But Brad, are you forgetting about BWS? How can you insist that it has to be applied to PRT but not to Robocars? Is their automatiom better than PRT?

Jack Slade

--- On Tue, 11/16/10, Brad Templeton <bra...@gmail.com> wrote:

> From: Brad Templeton <bra...@gmail.com>
> Subject: [t-i] On robocars, congestion and road capacity

Dennis Manning

non lue,
16 nov. 2010, 02:00:3516/11/2010
à transport-...@googlegroups.com
I agree. Lots of intermediary steps. That's why it will take so long. There
are a dizzyingly amount of ifs ands and buts on the path to blending what
the driver does versus what the computer does. Imagine the testing time
required to sort it out. This stuff doesn't lend itself to Moore's law. You
are dealing with humans and a massive in place physical system that you want
to incrementally change. Good luck. It's time to begin a separate new
network. It's simply not worth the effort, expense, and time to morph the
old system.

Given enough time AT&T could probably morph the old copper network. Given
enough time railroads could morph into cars. It's just that the intermediate
steps aren't worth the time and effort.

PRT is delivering the advantages now, not decades from now.

You of all people should grasp the dynamics of a disruptive technology. From
your early posts and your turning away from being a PRT enthusiast to
robocars I'm surprised that you have chosen the slow evolutionary approach
versus the small disruptive technology that can grow rapidly.

I have to conclude that you think robocars can get there first. I'm in
profound disagreement, but of course only time will tell.

Dennis

--------------------------------------------------
From: "Brad Templeton" <bra...@gmail.com>

Sent: Monday, November 15, 2010 9:25 PM

Brad Templeton

non lue,
16 nov. 2010, 02:46:1916/11/2010
à transport-innovators
What happens first and why is explained in detail on my web site so I
won't repeat it here. It's mostly based on who buys the tech and what
that drives developers to do. Moore's law keeps making all the
hardware cheaper and more powerful; what was expensive and hard to do
becomes cheap and easy, be it machine vision, or LIDARs or networking
other things.

The key is to not be trying to change the system but to do it from the
bottom up. Somebody buys a car with an autopilot in a place where
that's legal. Somebody subscribes to a traffic service. Some
supplier makes it better.

The ITS stuff -- roads with metering, congestion charging, redirection
-- that is the stuff that's much slower to come because governments
get involved. Almost as slow as PRT, though not nearly that slow
because there is nothing to build, just a few computers to install and
turn on if they get convinced it's a good idea. But yes, much slower
that the ground-up tech adoption. That gives us the single person
cars, the shorter headways, the load balancing of roads based not on
government issues reservations but the natural desire to take roads
with lower predicted usage.

Some government stuff comes faster. The street parking that's emptied
at rush hour already exists on major arterials -- the presence of
robocars makes it easier for cities to demand this and enforce it.
The traffic light timing and broadcasting of light times would be slow
if it were not already well in the works, along with a few other ITS
things.

PRT is not, alas "delivering these advantages now." There is no PRT
in a real city. There is just the argument that, should a city put in
some decent PRT, it could get short and predictable trip times for
commuters. And that might even be true, though it would have to be a
larger PRT than is likely to be built for quite a number of years.

On Nov 15, 11:00 pm, "Dennis Manning" <john.manni...@comcast.net>
wrote:
> I agree. Lots of intermediary steps. That's why it will take so long. There
> are a dizzyingly amount of ifs ands and buts on the path to blending what
> the driver does versus what the computer does. Imagine the testing time
> required to sort it out. This stuff doesn't lend itself to Moore's law. You
> are dealing with humans and a massive in place physical system that you want
> to incrementally change.  Good luck. It's time to begin a separate new
> network. It's simply not worth the effort, expense, and time to morph the
> old system.
>
> Given enough time AT&T could probably morph the old copper network. Given
> enough time railroads could morph into cars. It's just that the intermediate
> steps aren't worth the time and effort.
>
> PRT is delivering the advantages now, not decades from now.
>
> You of all people should grasp the dynamics of a disruptive technology. From
> your early posts and your turning away from being a PRT enthusiast to
> robocars I'm surprised that you have chosen the slow evolutionary approach
> versus the small disruptive technology that can grow rapidly.
>
> I have to conclude that you think robocars can get there first. I'm in
> profound disagreement, but of course only time will tell.
>
> Dennis
>
> --------------------------------------------------
> ...
>
> read more »

WALTER BREWER

non lue,
16 nov. 2010, 11:00:3216/11/2010
à transport-...@googlegroups.com
More broadly perhaps, this brings up the issue of "proven technology"
leaders use to turn aside significanly new approaches like PRT, Robocars.

Considering that most community leaders are on balance anti-automobile,(aka
coerce people out of cars), how will consideration of the PRT and Robocar
approaches compare?

Walt Brewer

Dennis Manning

non lue,
16 nov. 2010, 11:15:2616/11/2010
à transport-...@googlegroups.com
While Moore's law is running full tilt in the computer world the
transportation world moves rather slowly. You and I can endlessly debate
how the future for robocars and PRT will unfold. In the meantime we can both
monitor the progress. Only time will yield the results. As Yogi Berra says,
"Making predictions is hard. Especially when you are talking about the
future".

--------------------------------------------------
From: "Brad Templeton" <bra...@gmail.com>

Sent: Monday, November 15, 2010 11:46 PM

>> read more �

Jerry Roane

non lue,
16 nov. 2010, 11:36:4816/11/2010
à transport-...@googlegroups.com
Brad

Your approach does not address pollution or energy shortages coming.  Unless we make a revolutionary leap to the new paradigm we will be toast.  Said another way robocars alone equal toast.

What robo solution do you have for the extreme energy inefficiency of roads and liquid fuel engines?  The best I can think of would be to faring out a Nissan Leaf but then you are still leaving 50% on the table and with nuclear fuel good for 70 years what the hell are we going to do in year 71?  I will be dead of course so some would say why should I care?  Robocars without solving energy is simply saying you don't care about year 71.  You only care about your precious essay you wrote and feel like you need to continue to defend.  Time to evolve to address a bigger picture.  Advanced rolling stock running robo would be fine in my book but encouraging the Hummer and crossover liquid fuel cars in my opinion is a huge gaff.

Traffic will keep growing so the robo years are few, mostly because they would start way behind the eight ball.  The gains robo alone would do may just catch us up then as traffic expands with population explosion it will be engulfed again.  I agree we can drive better but I do not think the total improvement will be enough.  3% of drivers are enough to choke down a highway near the choke point.  Those same 3% will be on the highway with your robocars unless we create legislation to get those idiots off the road.  I would favor a performance based ejection program where based on your driving (robo monitoring) you either shape up or ship out as a human driver.  Reeducating these 3% drivers would drastically increase capacity with a mix of robocars and/or human driven cars.  Perhaps you might want to write an essay on robo monitoring using all the same techniques for robo driving but just as a human driving critic.  If you can take on this idea for a few days and digest it as a method toward robocars running advanced non-polluting all-solar cars then we might have some gain here from your thought processes.  It is difficult to open a discussion up for feature enhancement over conclusion advocacy but I suggest this little twist on the robocar idea as an evolutionary step that gets the roads working a few percent better and improves the weakest 3% of the driving public.  Computers certainly can improve our lives and sensors for traffic and exact lane positioning are here.  Using a pre-robocar approach where we coach the worst drivers would be of great value to society.  Just chew this one over a while and see what you think for a next step.   

Jerry Roane  




--

Brad Templeton

non lue,
16 nov. 2010, 13:12:0516/11/2010
à transport-innovators
I answer that in great detail, just not in the essay on congestion.

Small electric cars, suitable for the majority of trips (which are
short and urban) are vastly more efficient than today's cars, and
today's transit systems -- they are even more efficient than New York
subways and Japanese trains. That is in fact one of the big
arguments for PRT, in that it moves people in small electric
vehicles. However, there's a lot more to it than that. I started
this thread for the congestion essay, because a common question asked
here (which, in spite of the name of the newsgroup, is mostly PRT
people) is how to deal with congestion and capacity issues if you
don't have a big new network of guideways.
> ...
>
> read more »

eph

non lue,
16 nov. 2010, 14:04:3816/11/2010
à transport-innovators
I think we agree on small electric vehicles, but I want them now, not
in 50 years.

Small, electric, automated vehicles are available now in the PRT
form. Nothing else is available as public transit, so those
discussions do take a sideline.

Every time someone mentions a LRT project, I want to hear that
"robocars" are coming, can solve the congestion problem and that we
should hold off on building tracks. What kind of response do you
expect from the LRT crowd? How long will it take to automate BRT
which runs on dedicated roads? Even that is in the (distant) future.

PRT is here NOW. It's better than LRT and is small electric car
compatible (even automated ones, when they become available). LRT is
not.

F.
> ...
>
> read more »

Bruce Attah

non lue,
16 nov. 2010, 14:15:0416/11/2010
à transport-innovators
Personally, I'm not inclined to bet on robocars getting there first.
Moore's law does not apply to artificial intelligence. The hardware we
today have is probably already good enough to build fully autonomous
robocars, if only we could specify the problems well enough, figure
out the right algorithms/heuristics to solve those problems, and plug
them all together so they worked smoothly, and then test the result
thoroughly enough to ensure against any nasty surprises. The urban
street and the country road present a huge variety of hazards which
may crop up in a huge variety of circumstances. We can't say that if
we're 1/4 of the way to solving the problem today, we'll be 1/2 way in
18 months' time, and the whole way in 36 months. I think the rate of
progress is inherently unpredicatable. Furthermore, even if we had
what appeared to be a complete solution today, bringing it to market
would, I'm fairly sure, take several years, because of the range of
hazards and circumstances that the testing would have to cover before
anyone would insure the vehicle

I'm not convinced that nearly-autonomous vehicles will be very useful.
They might actually cause a lot of problems. If the vehicle seeks
human advice once every few minutes, but drives itself the rest of the
time, then the human operator's attention is apt to drift away. They
might start daydreaming, or they might get engrossed in a phonecall,
ebook, video game, spreadsheet, conversation, snack, makeup mirror, or
any of countless other distractions. When the vehicle seeks advice,
then, it will probably take several seconds before there's a response
-- by which time, the vehicle has stopped, and a traffic queue has
begun to form.

If robocar-topia is where we're going, then I suspect that we'll find
our way there fastest by the route that ULTra and 2getthere are
taking, since there's nothing to stop them, once the technology is
ready, from enabling their vehicles to drive off the guideway and onto
public roads.
> ...
>
> read more »

Jerry Roane

non lue,
16 nov. 2010, 15:07:2916/11/2010
à transport-...@googlegroups.com
Brad

Lets go one step deeper into the small electric car.  First is it necessary that the small electric car be small?  If a large (long) car is more efficient than a short and stubby cliche of a car wouldn't that be preferable?  It isd current myth that electric cars are short and stumpy yet the cars that placed and/or won in the progressive X Prize race were not tall and stumpy.  Just working on that myth.  Instead of short perhaps you meant energy efficient rather than define an attribute that may or may not be associated with energy efficient.  

Second if a car is optimized for only one speed won't it have an energy advantage over a "urban" electric car that has to shift gears and have a transmission or weak hill climbing ability if there is no transmission?  Again chipping away at the mythology.  Cars are cars.  There is nothing urban or country about a car.  It is what it is.  a car occupies the space of itself plus a safety buffer zone for reaction time.  If you make a car two feet narrower then the envelope in traffic is only two feet less and percentage-wise it is basically the same footprint in traffic flow.  A two second following distance defines a following buffer of 220 feet at 75 mph.  If the car is 12 feet long or if the car is 15 feet long the envelope in freeflow highway traffic is either 232 or 235 feet a difference of 1%.  Highway lanes are 14 feet wide so taking 3 feet out of a car to make it prone to rollover only takes the driving envelope from 14 feet to 12 feet or 17%.  We need a 15X not some fractional increment (17 or 1%) for the next paradigm.   This small car myth needs to be exploded.  it is simply not true and detracts from adoption because the myth is if my fat ass has to sit in a small car I will look silly by comparison to my ride.  Just the myth.  

Last myth "if you don't have a big new network of guideways."  Just build a nice network of gridded guideways.  duh!  It will be cheaper than continued highway building and emergency helicopter ambulance service.  Ever priced an overhaul on a helicopter big enough to hold an EMS worker and a paying customer?  The prices for operating EMS by helicopter is enormous.  Guideway cars would be far faster than a helicopter in a real emergency where your loved one needed fast medical care.

Jerry Roane 

> ...
>
> read more »

Brad Templeton

non lue,
16 nov. 2010, 15:14:0116/11/2010
à transport-innovators
Light rail fans believe in light rail. Facts, don't alter their
opinions, let alone future technologies. Robocar fans may be
optimistic about those vehicles, but 50 years? I'll take that bet.

PRT is a strange beast. You say it's hear now, but it isn't. Yes,
if a city wants to order a system, they can order one. That has been
true for many years. If your definition of "here now" is that there
are companies ready to fill orders, then what this means is that "here
now" is not a useful metric, since it does not predict whether cities
will have PRT.
> ...
>
> read more »

WALTER BREWER

non lue,
16 nov. 2010, 15:23:5416/11/2010
à transport-...@googlegroups.com
Fine. But there is a higher priority for robotics!
 
A robotic pat down machine for airport security.
 
Unisex preferred.
 
 Walt Brewer
----- Original Message -----

Brad Templeton

non lue,
16 nov. 2010, 15:36:0816/11/2010
à transport-innovators


On Nov 16, 11:15 am, Bruce Attah <bruce.at...@googlemail.com> wrote:
> Personally, I'm not inclined to bet on robocars getting there first.
> Moore's law does not apply to artificial intelligence. The hardware we

That statement is almost completely incorrect. In fact, for many of
the AI problems (which is different from the problem of AGI which
remains far away) it has been Moore's law and almost nothing but
Moore's law that has enabled the technologies you see today in speech
recognition, computer vision, translation, game playing and many other
fields. I don't want to give the algorithm guys too much short shrift
and they have done great stuff but the hardware has done more.

Fortunately, driving doesn't require AGI or anything close to it.
Hell, people even do it unconscious sometimes.


> today have is probably already good enough to build fully autonomous
> robocars, if only we could specify the problems well enough, figure
> out the right algorithms/heuristics to solve those problems, and plug
> them all together so they worked smoothly, and then test the result
> thoroughly enough to ensure against any nasty surprises. The urban
> street and the country road present a huge variety of hazards which
> may crop up in a huge variety of circumstances. We can't say that if
> we're 1/4 of the way to solving the problem today, we'll be 1/2 way in
> 18 months' time, and the whole way in 36 months. I think the rate of
> progress is inherently unpredicatable.
Perhaps but the way it goes is not on the timetable you describe.
Rather with exponential technologies you are 1/1000th of the way the
first year, 1/500th the 2nd
year, half way there the 10th year and all the way there the 11th
year.

That's not an exact prediction, but a description of past
technological trends in other computer driven fields.

Driving is not a trivial problem, but nor is it one of the great and
completely incomprehensible mysteries. It's mostly two problems,
"find your way" and "don't hit stuff." and do a better job than people
do at this.

That people are not that good at it (due to their own inattention and
other flaws robots don't have) is what makes it tractable.

><Furthermore, even if we had
> what appeared to be a complete solution today, bringing it to market
> would, I'm fairly sure, take several years, because of the range of
> hazards and circumstances that the testing would have to cover before
> anyone would insure the vehicle
>
> I'm not convinced that nearly-autonomous vehicles will be very useful.
> They might actually cause a lot of problems. If the vehicle seeks
> human advice once every few minutes, but drives itself the rest of the
> time, then the human operator's attention is apt to drift away. They
> might start daydreaming, or they might get engrossed in a phonecall,
> ebook, video game, spreadsheet, conversation, snack, makeup mirror, or
> any of countless other distractions. When the vehicle seeks advice,
> then, it will probably take several seconds before there's a response
> -- by which time, the vehicle has stopped, and a traffic queue has
> begun to form.

Seeking human advice that regularly is not something that anybody
would suggest is a suitable level of service. My personal view is
that you must never
need human advice for any urgent safety matter, but that it is OK to
seek it rarely for other matters. That the passenger can daydream is
a feature, and to be planned, not a bug.

But how often in your driving day do you encounter something that you
have never seen before, and must use your reasoning powers to figure a
solution? My impression is that this is more a once a week thing, if
that. But I am very interesting in compiling lists of unusual
traffic situations that people have encountered that they think could
present problems. The two main ones are unusual weather and debris
on the road. But because these are so well known, naturally teams
are going to spend a lot of time working out good solutions to those
problems. Computer systems are already a big part of making people
able to drive better on icy roads, not that people do it so well. I
am optimistic on this, but I also know that if it is not yet solved,
the solution is easy -- people still do the driving in snowstorms --
and that still leaves a big market.

Debris is also something we've collectively seen a lot of and will be
able to do a lot of testing on in advance. In many cases the answer
is simple -- drive around the debris. If cases where that can't be
done, it makes sense to call upon a human. But how often in your
driving day to you have to drive over substantive debris? (ie. not
light stuff that blows in the wind which indeed will involve work to
assure is properly identified.) One promising thing in CV is that
while a generalize CV system that can identify anything is hard,
specific CV systems to identify specific things (like cardboard, trash
bags, leaves, road kill and other common forms of debris) are much
easier.

I welcome contributions to the lists of things encountered on the road
that require special action.
> ...
>
> read more »

Dennis Manning

non lue,
16 nov. 2010, 16:57:4816/11/2010
à transport-...@googlegroups.com
Brad wrote (portion):


<Driving is not a trivial problem, but nor is it one of the great and
completely incomprehensible mysteries. It's mostly two problems,
"find your way" and "don't hit stuff." and do a better job than people
do at this.

I'd add a third "don't let stuff hit you". It might be the most difficult
one.

Bruce Attah

non lue,
16 nov. 2010, 18:14:1616/11/2010
à transport-innovators


On Nov 16, 8:36 pm, Brad Templeton <brad...@gmail.com> wrote:
> On Nov 16, 11:15 am, Bruce Attah <bruce.at...@googlemail.com> wrote:

>
> > Personally, I'm not inclined to bet on robocars getting there first.
> > Moore's law does not apply to artificial intelligence. The hardware we
>
> That statement is almost completely incorrect.  In fact, for many of
> the AI problems (which is different from the problem of AGI which
> remains far away) it has been Moore's law and almost nothing but
> Moore's law that has enabled the technologies you see today in speech
> recognition, computer vision, translation, game playing and many other
> fields.  I don't want to give the algorithm guys too much short shrift
> and they have done great stuff but the hardware has done more.

I think maybe I wasn't clear enough. You're quite right that hardware
improvements have played a critical role in the progress we've seen in
useful applications of AI over the past few decades, and hardware
improvements obey Moore's law. But that was my point precisely:
progress in developing heuristics and algorithms of the kind we call
"AI" is much less predictable. So far, it's been frustratingly slow.
Unlike hardware, it doesn't appear to obey Moore's law.


> Fortunately, driving doesn't require AGI or anything close to it.
> Hell, people even do it unconscious sometimes.

Generally, it's the stuff that people do unconsciously that's most
difficult for computers to replicate. Computers have been playing
chess at Grandmaster level for years, but choosing a suit of clothes
from a wardrobe and putting them on is something no robot has yet
achieved.
> ...
>
> read more »

eph

non lue,
16 nov. 2010, 20:03:2216/11/2010
à transport-innovators
I'm not saying it won't be possible before 50 years, just that
industry will try to maximize return on EVERY innovation every year or
worse, every vehicle refresh cycle which is every few years. Then
there is the pesky liability issue.

Moore's "law" may actually be Moore's "how to maximize profits
agreement" and if automated vehicle development follows similar
"laws"...

Unless google does something the auto industry will be blindsided by,
then it changes things. It's not in the auto industry's interest to
have cars that can be easily shared.

You know how the car was around for a long time before Henry Ford
figured out how to make them affordable then everyone wanted one?
People were scared of them, they were not horse drawn carriages. PRT
is in a similar infancy (it's not LRT). It takes time for people to
warm up to a new idea. There is a good chance that PRT/podcars will
hit that tipping point and systems will be installed in many places,
at least where overpriced LRT systems are being considered.

F.
> ...
>
> read more »

Jerry Schneider

non lue,
16 nov. 2010, 20:32:4316/11/2010
à transport-...@googlegroups.com
At 05:03 PM 11/16/2010, you wrote:
>I'm not saying it won't be possible before 50 years, just that
>industry will try to maximize return on EVERY innovation every year or
>worse, every vehicle refresh cycle which is every few years. Then
>there is the pesky liability issue.
>
>Moore's "law" may actually be Moore's "how to maximize profits
>agreement" and if automated vehicle development follows similar
>"laws"...
>
>Unless google does something the auto industry will be blindsided by,
>then it changes things. It's not in the auto industry's interest to
>have cars that can be easily shared.
>
>You know how the car was around for a long time before Henry Ford
>figured out how to make them affordable then everyone wanted one?
>People were scared of them, they were not horse drawn carriages. PRT
>is in a similar infancy (it's not LRT). It takes time for people to
>warm up to a new idea. There is a good chance that PRT/podcars will
>hit that tipping point and systems will be installed in many places,
>at least where overpriced LRT systems are being considered.

Is it wise to always tag LRT as being PRT's main
competitor? Isn't it a better strategy
to market PRT as a way to assist LRT (existing or
proposed) to be more attractive and
hence more heavily used? The same should be true for Express BRT service.

But, as long as conventional transit people
perceive (or are told) that PRT is going to diminish or
make them obsolete, is their any chance that they
could become advocates for PRT?
Is there any chance that PRT will be built without their advocacy?

Overpriced LRT systems are extremely popular with
cities as shown by the fact that
cities (pushed by their consultants) keep asking
the FTA for 50% cost-sharing support (i.e.
demand is much greater than the supply of federal
funds -- and has been since the FTA's New
Starts program has been in place (something like 20-30 years now).

>--
>You received this message because you are
>subscribed to the Google Groups "transport-innovators" group.
>To post to this group, send email to transport-...@googlegroups.com.
>To unsubscribe from this group, send email to
>transport-innova...@googlegroups.com.
>For more options, visit this group at
>http://groups.google.com/group/transport-innovators?hl=en.


- Jerry Schneider -
Innovative Transportation Technologies
http://faculty.washington.edu/jbs/itrans


Brad Templeton

non lue,
16 nov. 2010, 22:24:2316/11/2010
à transport-innovators
There are always people who feel threatened by any new technology
(even PRT) but there are lots of people who will sell it to you if
it's truly valuable. That's what a disruptive technology is. The
big boys don't understand it, and don't lead it out of fear of eating
their young. So they try to buy some government to stop it, and they
do succeed in slowing it down sometimes. But computer tech and
robotics are going to be different, I predict. If GM and Ford get the
U.S. governments to slow the robocars, other countries will embrace
them, and then we start seeing public perception that the US is
lagging the world in robotics, and as more people declare that
robotics is the new new thing for the coming decade, the harder it is
for GM and Ford to own the government on it.

At least for now though, many major car companies have active robocar
labs. Cars like Boss and Junior were sponsored by GM and Volkswagon.
At least some parts of these companies know it's coming, and they can
either lead it or become irrelevant. Amazingly, some of the big car
companies have realized that their last efforts to use the government
to protect themselves from competition didn't help them in the long
run.

Features in cars get adopted pretty quickly, especially in the luxury
end cars. How many years did it take before every car had a CD
player instead of a tape, or a bluetooth speakerphone, or a nav
system, at least as an option. Or Onstar or an mp3 player jack or
ABS. Or adaptive cruise. These techs are not universal, but they
went from minor to common in fairly short periods of time

But now consider a tech that can be improved with a firmware download,
or a motherboard or sensor replacement, rather than getting a new
car? I hope that the early vehicles deliberately try to be modular so
that it's easy to add new sensors if new sensors must be added, so
older vehicles can keep up.

I won't deny that many unconscious mental activities are not well
understood. The question is, do you have to solve them all like a
human does, and are there ones which we have no grasp of? As I go
down the risk of skills we're looking for, most of them are ones we
are getting better and better handles on (due to other market forces.)

Google has shown what you can do with excellent road data. As they
do this an interesting potential develops, which is cities embracing
the robocars. That means that they don't put up new signs or change
rules without updating the databases used by the robots. The robots
no longer have to be able to read and understand signs or at least
they don't have to ask a human very often. (Since our roads are
routinely driven by both foreign tourists and illiterates,
understanding all new signs is not a crucial safety thing in any
event. Yes, "bridge out" sounds scary but that is done with a
barrier, not just a sign.

So again, I hope to see people's lists of what they think are the big
hard problems. Particularly problems where you think a robot would
not know to stop and ask a human. I can think of silly ones -- two
men in mylar mirror suits carrying a plate glass window, a staple of
slapstick films -- but they are not real.

Not being hit is an important problem, though again the goal is just
to be better than people at it, and they have their issues. In most
cases, faster reaction times can give machines the edge. Right now I
would give humans the edge at understanding things like a deer at the
side of the road tensing to jump out. But there are only so many
species of megafauna and I think that teams dedicated to making
algorithms to work on specific problems can solve specific problems.
The hard thing, the thing humans do better, is solving general
problems. A robot might not know what everything that comes to the
side of the road is, but it can know that there is something that came
up to the side of the road. There will obviously be special purpose
systems for all types of people, all known vehicles and all major
animals, plus all sorts of classes of debris and blowing stuff. It's
a finite set, and not an AGI problem.
> ...
>
> read more »

eph

non lue,
16 nov. 2010, 22:28:3716/11/2010
à transport-innovators
You're right from a sales perspective (maybe), but from a user
perspective, why would you leave a personal podcar to take an
overcrowded bus or LRV? Or leave a podcar to wait for an LRV that
might come in 30 minutes if it's still running late in the evening?

From a city perspective, why would you install an expensive money-
losing system when there is a alternative that could at least cover
it's O&M? Why would you add to congestion by removing 3 vehicle lanes
down a main corridor or rows of parking? If that's a goal, there are
cheaper ways of achieving this than with LRT.

This is fear of the LRT lobby and propaganda machine (not misplaced)
and isn't right (thought it is reality). People have to wake up and
see what is real and what is a slick sales job. People have to learn
and see what the truth is and demand it from their gov't.

I think the LRT lobby knows PRT is a threat, otherwise there wouldn't
be such an effort to kill any PRT prospects. People can be fooled for
a while, but eventually they realize what the game is and I know that
not all city workers can be bought/fooled. Some really care about
their city and want what's best. They need some support and a little
leadership.

BTW. LRT means less drivers are needed - that's one of their big
selling points. At least PRT should attract ridership from the car
driving side instead of just replacing daytime bus service. In Paris,
buses run instead of the metro lines at night. With PRT, night shifts
are reduced and day shifts are still there.

F.
> ...
>
> read more »

Dennis Manning

non lue,
17 nov. 2010, 02:06:5217/11/2010
à transport-...@googlegroups.com

--------------------------------------------------
From: "Brad Templeton" <bra...@gmail.com>

Sent: Tuesday, November 16, 2010 7:24 PM


To: "transport-innovators" <transport-...@googlegroups.com>
Subject: [t-i] Re: On robocars, congestion and road capacity

> There are always people who feel threatened by any new technology


> (even PRT) but there are lots of people who will sell it to you if
> it's truly valuable. That's what a disruptive technology is. The
> big boys don't understand it, and don't lead it out of fear of eating
> their young. So they try to buy some government to stop it, and they
> do succeed in slowing it down sometimes. But computer tech and
> robotics are going to be different, I predict. If GM and Ford get the
> U.S. governments to slow the robocars, other countries will embrace
> them, and then we start seeing public perception that the US is
> lagging the world in robotics, and as more people declare that
> robotics is the new new thing for the coming decade, the harder it is
> for GM and Ford to own the government on it.

Were people threatened by digital watches? I'd say only those a with stake
in mechanical watches. Lot's of people will sell them to you if it's
valuable? Not usually. Those pushing a disruptive technology are often
confined to relatively few producers? Good remark about the big boys. In
some cases they don't understand the new technology, but I think more often
it's because because it threatens what they do best, or where they are
making the most profit. It's all explained rather well in the "Innovator's
Dilemma". AT&T and the TV industry have clearly used the gov't to slow the
onslaught of the Internet. They are losing. De ja vu. The telegraph people
did the same thing to the infant phone industry.


>
> At least for now though, many major car companies have active robocar
> labs. Cars like Boss and Junior were sponsored by GM and Volkswagon.
> At least some parts of these companies know it's coming, and they can
> either lead it or become irrelevant. Amazingly, some of the big car
> companies have realized that their last efforts to use the government
> to protect themselves from competition didn't help them in the long
> run.

More than robocars the GMs are coming late to the electric car party.

>
> Features in cars get adopted pretty quickly, especially in the luxury
> end cars. How many years did it take before every car had a CD
> player instead of a tape, or a bluetooth speakerphone, or a nav
> system, at least as an option. Or Onstar or an mp3 player jack or
> ABS. Or adaptive cruise. These techs are not universal, but they
> went from minor to common in fairly short periods of time

Adopted fairly quickly??? It's easy to name a lot of improvements that were
very slow. Detroit drug it's feet on a lot of things - radial tires, fuel
injection, front wheel drive, etc. There's a lot of time of adoption
differences for different improvements. Some are fast. Some take a long
time.

>
> But now consider a tech that can be improved with a firmware download,
> or a motherboard or sensor replacement, rather than getting a new
> car? I hope that the early vehicles deliberately try to be modular so
> that it's easy to add new sensors if new sensors must be added, so
> older vehicles can keep up.

Talk about time consuming!

>
> I won't deny that many unconscious mental activities are not well
> understood. The question is, do you have to solve them all like a
> human does, and are there ones which we have no grasp of? As I go
> down the risk of skills we're looking for, most of them are ones we
> are getting better and better handles on (due to other market forces.)

Better handles? Perhaps, but what a long way to go.


>
> Google has shown what you can do with excellent road data. As they
> do this an interesting potential develops, which is cities embracing
> the robocars. That means that they don't put up new signs or change
> rules without updating the databases used by the robots. The robots
> no longer have to be able to read and understand signs or at least
> they don't have to ask a human very often. (Since our roads are
> routinely driven by both foreign tourists and illiterates,
> understanding all new signs is not a crucial safety thing in any
> event. Yes, "bridge out" sounds scary but that is done with a
> barrier, not just a sign.

What can you do with road data? What difference has it made so far? The bulk
of drivers don't get any road data other than a few changeable message signs
on major freeways. Telling me there's a wreck ahead on the freeway isn't
useful unless I get it in time and there is a better alternate and knowing
how to use the alternate.

>
> So again, I hope to see people's lists of what they think are the big
> hard problems. Particularly problems where you think a robot would
> not know to stop and ask a human. I can think of silly ones -- two
> men in mylar mirror suits carrying a plate glass window, a staple of
> slapstick films -- but they are not real.

The list would be so long it wouldn't be worth my time. Let alone the time
to figure out the software to mitigate the circumstance.

>
> Not being hit is an important problem, though again the goal is just
> to be better than people at it, and they have their issues. In most
> cases, faster reaction times can give machines the edge. Right now I
> would give humans the edge at understanding things like a deer at the
> side of the road tensing to jump out. But there are only so many
> species of megafauna and I think that teams dedicated to making
> algorithms to work on specific problems can solve specific problems.
> The hard thing, the thing humans do better, is solving general
> problems. A robot might not know what everything that comes to the
> side of the road is, but it can know that there is something that came
> up to the side of the road. There will obviously be special purpose
> systems for all types of people, all known vehicles and all major
> animals, plus all sorts of classes of debris and blowing stuff. It's
> a finite set, and not an AGI problem.

No, you have to be much better than humans as we have learned when safety
concerns are raised about PRT. Bashing cars like LRT does isn't anywhere
near acceptable for PRT. To be accepted PRT will have to be many times safer
than LRT or human controlled cars, and so will robocars. As far as reacting
properly to a pedestrian or animal or whatever is near the streets edge
about the only robocar response to the unpredictable action is to go very
slow or stop thereby inviting to be rear ended. If the robocar senses the
car to the rear is close perhaps the best decision is to take the chance
that the pedestrian/animal will stay put. That's not a great prospect for
robocar safety requirements. Lot's of car wrecks have taken place when
people hit the brakes or swerve to avoid animals. Robocars might be able to
make a quicker more odds on safe decision like go ahead and whack Fifi. I
wonder if the Robocar would stop or do a hit and run? In a brief stretch a
few years ago between the wife and I we had 4 accidents where were hit and
we weren't even moving. The point is that in a road environment even a
perfect driver can't avoid accidents traveling in mixed traffic. So I ask -
why subject near perfect driving machines to a very imperfect road
situation. I think separating them from existing manual drivers is more
important than trying to make use of existing roads. In other words a really
disruptive approach rather an incremental one.

Just thought I'd present another difficulty for robocars. When you are on a
one lane each way highway and the oncoming car is drifting your way at what
point does the robocar decide the oncoming car is presenting a danger? and
how does it react? It might move to a shoulder, but what if there's little
shoulder? or if the shoulder condition is changing abruptly a few feet
ahead. What if? What if? What if? Too damn many what ifs?

>> read more �

Kirston Henderson

non lue,
17 nov. 2010, 02:28:0817/11/2010
à transport-...@googlegroups.com

On Nov 17, 2010, at 1:06 AM, Dennis Manning wrote:

>>
> What can you do with road data? What difference has it made so far?
> The bulk of drivers don't get any road data other than a few
> changeable message signs on major freeways. Telling me there's a
> wreck ahead on the freeway isn't useful unless I get it in time and
> there is a better alternate and knowing how to use the alternate.
>>
> The list would be so long it wouldn't be worth my time. Let alone
> the time to figure out the software to mitigate the circumstance.
>
If you simply transport both current and future, possibly short-range
electric cars recharged during guideway travel, on automated
CarFerries™ that move at much higher speeds than road traffic, you
have solved most of the problem without the great expense and risk of
robocars. Automated CarFerries moving on elevated guideways present a
far less complex development task. That development task is almost
precisely the same as it is for PRT. In our own case, we are
reasonably sure that that even if we start operations with manually-
controlled, coupled trains, we will probably have reached the point of
being able to fully automate operation to individually dispatched
vehicles by the time that we are able to install a significant amount
of guideway.

By the way, the task of automating a guideway vehicle is much less
complex because you are automating only a single axis and have far
fewer things to sense than if you attempt to build a robocar.

Kirston Henderson
MegaRail® Transportation Systems, Inc.

Jack Slade

non lue,
17 nov. 2010, 02:54:0217/11/2010
à transport-...@googlegroups.com

--- On Wed, 11/17/10, Jerry Schneider <j...@peak.org> wrote:
>
> Is it wise to always tag LRT as being PRT's main
> competitor? Isn't it a better strategy
> to market PRT as a way to assist LRT (existing or proposed)
> to be more attractive and
> hence more heavily used? The same should be true for
> Express BRT service.<<<<

You are probably right about that, Jerry. It is just that, on this list, we sometimes forget that wen are not just talking to each other. There are lots of people monitoring who do not join in and make postings, and a search will turn up postings we did 5 years ago or more. We just forget.

Personally, I think a system serving as a feeder for a train or LRT system would be a great start. I would even like to provide service for passengers when they are leaving the ststions, all I want is the chance.
What John Q Public might think of this is just conjecture, better left unpublished.

Jack Slade

Brad Templeton

non lue,
17 nov. 2010, 03:19:1817/11/2010
à transport-innovators
Note that as we've had other robocar discussions here, my main purpose
in this thread was to discuss the implications of robots and ITS on
increasing road capacity and reducing congestion.


> Adopted fairly quickly??? It's easy to name a lot of improvements that were
> very slow. Detroit drug it's feet on a lot of things - radial tires, fuel
> injection, front wheel drive, etc. There's a lot of time of adoption
> differences for different improvements. Some are fast. Some take a long
> time.
>

Yes, but when the public really wants it, when it becomes a key
differentiator from other cars, it gets adopted in cars very quickly,
particularly at the high end.

Even if the legal use is constrained to subsets of streets, I am
comfortable saying people will really, really want this. It gives
them something so few innovations can -- time. Even more than safety.

But they'll buy the predecessor technologies -- accident avoidance
etc. -- for safety.



> What can you do with road data? What difference has it made so far? The bulk

It allowed Google's cars to take trips of up to 1,000 on crowded
streets without anything that would make the human supervisor want to
intervene, that's what road data did.


>
> Just thought I'd present another difficulty for robocars. When you are on a
> one lane each way highway and the oncoming car is drifting your way at what
> point does the robocar decide the oncoming car is presenting a danger? and
> how does it react? It might move to a shoulder, but what if there's little
> shoulder? or if the shoulder condition is changing abruptly a few feet
> ahead. What if? What if? What if? Too damn many what ifs?

A good system would be tracking the anomaly regularly, and plotting
when it might get close to a point where escape is not possible. If
the other car has not corrected as this point approaches, the car
would probably take actions to give it more options, such as slowing,
looking for places with shoulder. It might change velocity so that
it will pass the other car at a spot which does offer some shoulder.
It might flash lights at the other car.

Whatever you would do in this situation. I do agree that undivided
roads present a special problem in defensive driving that we just
ignore. On such roads, at any moment an oncoming driver can suddenly
veer and put you in a head-on, and the best driving can't prevent
it. We've chosen in our driving to just ignore that, and we may well
program our robots the same way.
>

Michael Weidler

non lue,
18 nov. 2010, 06:41:0618/11/2010
à transport-...@googlegroups.com
When you can show me how you will avoid deer without killing the passengers of the vehicle then there may be something to talk about.


--- On Tue, 11/16/10, Brad Templeton <bra...@gmail.com> wrote:

From: Brad Templeton <bra...@gmail.com>
Subject: [t-i] Re: On robocars, congestion and road capacity
--
You received this message because you are subscribed to the Google Groups "transport-innovators" group.
To post to this group, send email to transport-...@googlegroups.com.
To unsubscribe from this group, send email to transport-innovators+unsub...@googlegroups.com.

Michael Weidler

non lue,
18 nov. 2010, 06:51:3918/11/2010
à transport-...@googlegroups.com
If GM and Ford think robocars are bad for business, why wouldn't overseas car makers think in a similar fashion?


--- On Tue, 11/16/10, Brad Templeton <bra...@gmail.com> wrote:

From: Brad Templeton <bra...@gmail.com>
Subject: [t-i] Re: On robocars, congestion and road capacity
To: "transport-innovators" <transport-...@googlegroups.com>

Michael Weidler

non lue,
18 nov. 2010, 06:35:1218/11/2010
à transport-...@googlegroups.com
I want one which vibrates!!!!!

--- On Tue, 11/16/10, WALTER BREWER <catc...@verizon.net> wrote:

Brad Templeton

non lue,
18 nov. 2010, 15:29:3918/11/2010
à transport-innovators


On Nov 18, 3:51 am, Michael Weidler <pstran...@yahoo.com> wrote:
> If GM and Ford think robocars are bad for business, why wouldn't overseas car makers think in a similar fashion?
>
Only big automakers would decide a new technology would threaten their
business. The reason they would decide this is because small to
medium automakers were chomping at the bit to use the new technology
to threaten the big boys. A new positive technology will always
have its boosters and builders, and those who are threatened by it.
Fortunately, the boosters usually win, though they may be slowed for a
bit. Otherwise we wouldn't get anything new.

Right now though, the big car companies in the USA feel afraid for
their lives for other reasons, and are more open to innovation than
before. GM is repositioning itself as "general mobility" and is
publishing vision statements where robocars and services are its
future. Of course, in time they might switch to fear rather than
opportunity. But if they do, others will take their place, of that I
have little doubt.

> When you can show me how you will avoid deer without killing the passengers of the vehicle then there may be something to talk about.

Deer present a problem for human drivers and robotic driving.
Strategies are similar, though the robot has several advantages in
spotting and dealing with the deer:

a) The robot's IR-transmitting camera will light up the deer's eyes
like a beacon, even at dusk and at night. It is just after sunset
when the deer seem to be on the move and present the greatest problem
to human drivers. LIDAR sees equally well at night (even slightly
better.)

b) The robocar has a complete 3-D map of the road, including the
location of every bush, fencepost, ditch, curb, tree etc. If it
detects an object in the field that is not normally there, it will do
extra scrutiny and exercise extra caution. Extra scrutiny can include
both visual inspection, radar inspection, more detailed LIDAR, motion
detection etc.

c) While nobody is yet using them because they are expensive, far
infrared cameras which measure heat are great, particularly outside of
sunshine hours, at spotting animals, people and things with engines.
They can spot an animal in the brush far better than human eyes from
the heat signature.

d) If the deer steps into the road it will of course be spotted.
Today's LIDAR sees to 100m. I expect it to improve a bit, but at 100m
you can stop at 60mph. Above 60mph you can't brake in time but you
will impact at a slow speed. Fast reaction time makes the difference.

e) Fast reaction time also offers the ability to simply swerve around
the deer unless there is heavy traffic and oncoming traffic. Robots
can adjust their speed to know that they will encounter the obstacle
at a point where there is room to swerve. If the oncoming traffic is
also a robot it can be communicated with to assure this, otherwise you
just make a range of predictions about what it might do, but you will
judge the range of positions perfectly, your doppler is telling you
precisely every change of velocity of the other vehicle.

f) If, in spite of this, you are going to hit it, you can deploy a
front of car airbag just before hitting to soften the blow. You can
also choose to hit it with a section of the car designed to absorb the
shock. I could even imagine a special bumper which could be extended
quickly (compressed gas) in front of the car, said bumper mounted on
shock absorbing struts so that the energy is absorbed over a longer
distance. For a 200lb (typical) deer this might be quite effective.
This bumper might be tilted left or right to direct the deer away from
the car into the ditch, deflection transfers less momentum and energy
to the car. This is something new I just thought of, something that
wasn't possible before today's tech so I will be giving it some more
thought.

As noted, these are all things the robot can do that people can't.
The robot would also be doing most of what people can do. Today's
robocars however don't use vision to identify objects, and there is
not an existing CV system that does general identification of
arbitrary objects, and I can't tell you when such will come though I
think it will come. However, the 3D vision of LIDAR is in many ways
superior to human vision at spotting objects.

Jerry Schneider

non lue,
18 nov. 2010, 17:16:1918/11/2010
à transport-...@googlegroups.com
At 12:29 PM 11/18/2010, Brad wrote:

>Deer present a problem for human drivers and robotic driving.
>Strategies are similar, though the robot has several advantages in
>spotting and dealing with the deer:

An additional scenario: A friend of mine was killed by a collision
with a large deer
that tried to jump over his moving auto and crashed directly into the
windshield.


Brad Templeton

non lue,
18 nov. 2010, 18:44:0518/11/2010
à transport-innovators
Strategies for avoidance or for slowing down obviously would help with
this.

However, if you really felt this was a concern -- and I feel, aside
from the tragedy of your friend that it is not a frequent enough event
to have most people want it -- robocars have the option of having a
much stronger windshield. In fact, of not having a windshield at
all, though I think most people would prefer to be able to see forward
both for the view and to avoid motion sickness.

However, unlike human driven cars, the windshield does not have to be
a giant picture window, it could have panes, with strong bars between,
which would eliminate the risk you outline.

Some cars, like sleeper cars may not have a front windshield.
Instead, if the passenger needs to see the situation, the HDTV can
show it pretty well. But most people will want one.

Also possible, though expensive, is a plate that can come up from
below and cover the windshield in an emergency. I don't think the
number of injuries from things coming through the windshield is high
enough to justify this cost but I could be in error.

Jerry Roane

non lue,
18 nov. 2010, 21:17:0018/11/2010
à transport-...@googlegroups.com
Brad

Frequency of deer wrecks is pretty high.  I hit a deer with my last car and it was the deer damage that was the deciding factor on selling it off versus fixing it back up.  I was traveling 65 mph when I saw the deer and I was able to slam on the brakes and slow to about 30 mph before impact but the deer could have easily been lethal with a few feet one way or another.  No windshield can take a deer strike.  Windshields are lucky to take a bird of prey strike.  The stats on moose hits are higher than you might think too because of their high mass and large size.  Elevated guideway would avoid all these problems of large animals killing drivers and passengers.  

Since guideway is cheaper than asphalt it would be hard to justify switching from guideway to roads if it was reversed.  Who would pay more for less safety?  What came first was guideway as railroads predate cars on roads.  We are just suggesting to go back to the first idea and evolve it some more with 3D control of vehicle trajectory.  While we are at it we can solve your deer problem for free.  

Jerry Roane 

WALTER BREWER

non lue,
18 nov. 2010, 21:46:5618/11/2010
à transport-...@googlegroups.com
On some Interstates in Ohio, in deer infested areas, there are deer warning systems.
 
Don't know the radiation band, but emitters are aimed at detectors spaced perhaps 1/2 mile apart.
Interuption of the beam creates a warning signal.
 
I've never seen one in the detection mode.
 
 Walt Brewer
 
----- Original Message -----

Brad Templeton

non lue,
18 nov. 2010, 22:28:4018/11/2010
à transport-innovators
As I said, it's a trade-off. You would look at the risk from deer
and decide what makes sense to spend. Cars of the future (not just
robocars) will have the ability to know they are unavoidably going to
be in an accident some amount of time before the accident (half a
second to perhaps a couple of seconds.) Physics is something
computers do well. So if the cost is justified, you could have metal
rods that shoot up ever foot in front of the windshield driven by
compressed gas, so that little can get through. As I noted, the
robocar window can be much stronger than a driving window as it is
allowed to have visual obstructions while those are forbidden in human
driven cars.

There are a few other tricks that can be used, such as sensors on the
road that see the deer and report that to cars. This is one of those
Moore's law things -- today a solar powered LIDAR radio on a pole
might cost $50,000 but make it in quantity and it costs under $1,000
and that might be worth it, allowing robots and drivers to know if
deer are on the road, or crossing in groups, or interrupting laser
fences (that's pretty cheap, actually, and as noted already exists.)

While in theory one could build rural guideway systems, and I know
folks want to sell them, rural PRT is not on the near horizon, I would
judge.

On Nov 18, 6:17 pm, Jerry Roane <jerry.ro...@gmail.com> wrote:
> Brad
>
> Frequency of deer wrecks is pretty high.  I hit a deer with my last car and
> it was the deer damage that was the deciding factor on selling it off versus
> fixing it back up.  I was traveling 65 mph when I saw the deer and I was
> able to slam on the brakes and slow to about 30 mph before impact but the
> deer could have easily been lethal with a few feet one way or another.  No
> windshield can take a deer strike.  Windshields are lucky to take a bird of
> prey strike.  The stats on moose hits are higher than you might think too
> because of their high mass and large size.  Elevated guideway would avoid
> all these problems of large animals killing drivers and passengers.
>
> Since guideway is cheaper than asphalt it would be hard to justify switching
> from guideway to roads if it was reversed.  Who would pay more for less
> safety?  What came first was guideway as railroads predate cars on roads.
>  We are just suggesting to go back to the first idea and evolve it some more
> with 3D control of vehicle trajectory.  While we are at it we can solve your
> deer problem for free.
>
> Jerry Roane
>
> > transport-innova...@googlegroups.com<transport-innovators%2Bunsu...@googlegroups.com>
> > .

Jack Slade

non lue,
19 nov. 2010, 02:55:0819/11/2010
à transport-...@googlegroups.com
Walter, I have never heard of this. All I have ever seen is warning signs like "Deer Crossing", which doesn't mean much to a driver, especially at night. Does it make a roadside sign flash, or what? How would it react to Coyotes and other smaller animals?

Jack Slade

Michael Weidler

non lue,
19 nov. 2010, 03:11:3719/11/2010
à transport-...@googlegroups.com
Deer wind up in windshields all the time. Usually it is because they are flipped up there by the impact. Judging by your response to this and other posts, you apparently have not had much experience with deer. As I have stated before, most anything you do to avoid the deer is going to cause harm to the passengers.


--- On Thu, 11/18/10, Brad Templeton <bra...@gmail.com> wrote:
--
You received this message because you are subscribed to the Google Groups "transport-innovators" group.
To post to this group, send email to transport-...@googlegroups.com.
To unsubscribe from this group, send email to transport-innovators+unsub...@googlegroups.com.

Michael Weidler

non lue,
19 nov. 2010, 03:28:2019/11/2010
à transport-...@googlegroups.com
Unfortunately, those sensors never seem to reset themselves, so they are always "on".


--- On Thu, 11/18/10, WALTER BREWER <catc...@verizon.net> wrote:

Brad Templeton

non lue,
19 nov. 2010, 03:45:2419/11/2010
à transport-innovators
What experience do you want me to have with the deer? I have never
hit one. Seen plenty on the roads, though. Living 31 years in Canada
will do that to you. I don't agree that any moves to avoid the deer
are going to injure the passengers. I have never hit one because
I've avoided them many times, usually by braking and proceeding with
caution, which is what I presume most other people do. I do agree
the number of collisions is quite high, and I had not realized there
were 200 deaths each year. (That's a small fraction of total
traffic deaths, but still significant.)

It still seems to me that the robots have significant advantage here.
Better able to see the deer (especially in the dark), faster to
respond, and unlike humans who often panic and skid while trying to
avoid a deer, never panicking and always knowing what will skid and
what won't. Add the ability to have a windshield that the deer can't
penetrate, a bumper and external airbag to cushion the shock if there
has to be impact and some other things and I think the robots are
going to be the clear winner.

On top of that, it might be productive to do research into other
things the robots could do. Is there a sound that might startle the
deer and make it leave the road? Would shooting the deer with an
airsoft BB or similar cause it to bolt? If the headlights are
keeping the deer transfixed, the robocar can turn them off (they are
mostly there to keep the humans comfortable, since at present no
robocar uses headlights to see at night.) While I can't guarantee
it, it's not out of the question that research could discover a series
of things with light, sound and projectiles that would have a high
probability of spooking a deer. Now the projectile thing is pretty
far out, in general I don't see a need to arm the cars, even with
airsoft, but it's something that could be tried. Humans could
trigger these sorts of defences too, except humans can't afford to
turn off the headliights.

On Nov 19, 12:11 am, Michael Weidler <pstran...@yahoo.com> wrote:
> Deer wind up in windshields all the time. Usually it is because they are flipped up there by the impact. Judging by your response to this and other posts, you apparently have not had much experience with deer. As I have stated before, most anything you do to avoid the deer is going to cause harm to the passengers.
>
> --- On Thu, 11/18/10, Brad Templeton <brad...@gmail.com> wrote:
> To unsubscribe from this group, send email to transport-innova...@googlegroups.com.

Michael Weidler

non lue,
19 nov. 2010, 03:51:1319/11/2010
à transport-...@googlegroups.com
32 miles from Marengo to Pearsons near the Cedar Rapids Airport. Divide by . 6 to get km. Multiply by 1000 to get meters. Divide again by 100 to account for 100m range.  And that's 533 units or $533,000 for just ONE SIDE of ONE ROAD. Not to mention, it does not stop the deer. It just let's you know that one is about to run into you!!

Furthermore, if the beam does more than just provide an optical fence - for instance scans into the field or woods - then the alarm will be going off continually. If you interface this system with robocars, what happens to me when my robo chauffeur slams on the brakes because it's seeing deer everywhere?

--- On Thu, 11/18/10, Brad Templeton <bra...@gmail.com> wrote:

From: Brad Templeton <bra...@gmail.com>
Subject: [t-i] Re: On robocars, congestion and road capacity
To: "transport-innovators" <transport-...@googlegroups.com>
> > transport-innovators+unsub...@googlegroups.com<transport-innovators%2Bunsu...@googlegroups.com>

> > .
> > For more options, visit this group at
> >http://groups.google.com/group/transport-innovators?hl=en.

--
You received this message because you are subscribed to the Google Groups "transport-innovators" group.
To post to this group, send email to transport-...@googlegroups.com.
To unsubscribe from this group, send email to transport-innovators+unsub...@googlegroups.com.

Michael Weidler

non lue,
19 nov. 2010, 04:12:1619/11/2010
à transport-...@googlegroups.com
Brad,

Here is an experiment for you to try. Have your spouse (or someone) drive down the road at 60mph and randomly spike the brakes while your eyes are closed. The closed eyes simulates darkness.

Now if that doesn't convince you that robocars are a bad idea, try the follow up experiment. Run the car into a few bales of hay at 60mph (eyes closed again) to simulate actually hitting a deer at night.

What you can not seem to grasp is that when you are driving you are presumably paying attention to what is going on. You are aware that something is about to happen. If you are being chauffeured by robo car, you are NOT PAYING ATTENTION to what is happening around you. After all, that is the whole purpose of robo car. Therefore, you are not likely to be ready for any sudden stops or evasive maneuvers.

BTW, frightening deer is just as likely to make them bolt into you as make them run away.


--- On Fri, 11/19/10, Brad Templeton <bra...@gmail.com> wrote:
> To unsubscribe from this group, send email to transport-innovators+unsub...@googlegroups.com.

> For more options, visit this group athttp://groups.google.com/group/transport-innovators?hl=en.

--
You received this message because you are subscribed to the Google Groups "transport-innovators" group.
To post to this group, send email to transport-...@googlegroups.com.
To unsubscribe from this group, send email to transport-innovators+unsub...@googlegroups.com.

Brad Templeton

non lue,
19 nov. 2010, 04:34:3719/11/2010
à transport-innovators

I found the deer problem interesting enough that I summarized these
thoughts and some others into a blog post

http://ideas.4brad.com/robocars-vs-deer-and-flying-bumper

Brad Templeton

non lue,
19 nov. 2010, 04:46:3619/11/2010
à transport-innovators


On Nov 19, 1:12 am, Michael Weidler <pstran...@yahoo.com> wrote:
> Brad,
>
> Here is an experiment for you to try. Have your spouse (or someone) drive down the road at 60mph and randomly spike the brakes while your eyes are closed. The closed eyes simulates darkness.

Already done it, been in the car when the driver needs to brake
suddenly for a deer. It's quite scary of course. But I have my
seatbelt on, and I recommend it if you are going to drive in deer
crossing areas at high speeds!

Of course, for the passengers at the front facing backwards in the
robocar, it's not nearly so bad. Can even get away without the
seatbelt I suppose.

>
> Now if that doesn't convince you that robocars are a bad idea, try the follow up experiment. Run the car into a few bales of hay at 60mph (eyes closed again) to simulate actually hitting a deer at night.

Going to be even more shocking. Not interested in doing the
experiment. The robocar will of course trigger airbags at the right
moment if there is an impact, a bit better than today's cars do but
similar principle.
>
> What you can not seem to grasp is that when you are driving you are presumably paying attention to what is going on. You are aware that something is about to happen. If you are being chauffeured by robo car, you are NOT PAYING ATTENTION to what is happening around you. After all, that is the whole purpose of robo car. Therefore, you are not likely to be ready for any sudden stops or evasive maneuvers.

Yes, it will be an uncomfortable ride if you car has to make sudden
moves.
>
> BTW, frightening deer is just as likely to make them bolt into you as make them run away.
>
Well, that's what we would want to perform various experiments to
test. There may be no method which is reliable at getting them off
the road, but it's definitely worth looking into. Human drivers could
use some of those methods as well, though their problem is they
usually are in a panic when they see something on the road in front of
them and are slamming on the brakes or swerving. Robots may not have
brains the size of a planet, but they take the cover of the HHGTTG
very seriously.

WALTER BREWER

non lue,
19 nov. 2010, 10:01:5219/11/2010
à transport-...@googlegroups.com
I believe there is/was an installation east of Cleveland on I-90.

May have been experimental.

Yes I can see other significant size animals might trigger. I seem to
remember seeing a fence also.

Ohio DOT should know about it.

Jerry Roane

non lue,
19 nov. 2010, 10:14:4719/11/2010
à transport-...@googlegroups.com
Brad

No one is going to have a pole sticking out the top of their car.  What is the Cd of that thing?  I can see some guy in a pedestrian wreck being skewered by your deer spikes and that would be the end of that idea.  Those seem to be grasping a bit because you wish for a quick answer.  Nothing wrong with wishing for a solution to a problem that is the first step in all innovation a healthy discontent with what is available today.  A better solution than pedestrian killing spikes powered by stored gas pressure energy that will leak might be elevated roads or guideways.  Just a thought.

The last sentence is the most important.  How can we served rural locations or the outer band of populated areas.  Doing the math and depending on your definition of where the outer boundary of population area is there are more people (tax payers) on the outer band than all the inner ghettos of the nation.  It is just geometry.  Dual mode serves the outer band just fine so their tax dollars can easily go toward a democratic fair system if they get the benefit also.  Unlike commuter rail if you live East and the rail runs North/South then you pay the tax and get nothing.  If you live North but live too far North you still pay tax and get nothing.  A shared system using roads for feeders and guideway to movement and high speed rail or more likely airports for long distance then each tax payer gets essentially the same good value from that.  Rural areas just drive in from their outlying farm and use the city grid to avoid any traffic.  Rural roads do not have traffic other than the occasional farm implement or horse trailer.  With a few strategically placed passing lanes rural traffic is solved.  With elevated guideway city traffic is solved.  It can be robo or not robo you get to your destination just as fast and just as non-lethal.  If robo driving is to reduce fatalities and car travel is made non-lethal independent of robo driving then the gain of robo driving is dents in the fender occasionally (maybe) and the ability of the diver to in the mythology "take a nap or read a novel".  How many naps do we need and how many books of any importance do we read really?   

Jerry Roane 

To unsubscribe from this group, send email to transport-innova...@googlegroups.com.

WALTER BREWER

non lue,
19 nov. 2010, 10:15:3519/11/2010
à transport-...@googlegroups.com
A webbed frame that flips up upon deer detection might keep the deer from
the windshield.

But a properly designed airbag sounds more reasonable and practical. It
would be more multipurpose; small collisions, failure of electronic
tailgating on freeways, etc, etc.

Steve Raney territory.

Driven in deer country a lot, but been lucky. A good friend and wife had one
come through their windshield and land in laps.
Deer killed. They were pretty well cut up.

Walt Brewer

WALTER BREWER

non lue,
19 nov. 2010, 10:28:3419/11/2010
à transport-...@googlegroups.com
Another aspect of deer is their road crossing habits. I note with some amusement the deer warning signs that define the distanse ahead with great implied presision. Numbers like 1/2 mile. Sometimes down to 100's of feet.
 
 This suggests the more cost effective way statistically is to provide tunnels for the deer with modest fencing.
I suspect fencing alone won't work; they will move their paths to compensate.
I'm aware of doing this when new roads cross cow pastures, and it works. Deer are probably smarter than cows.

Michael Weidler

non lue,
19 nov. 2010, 11:50:5019/11/2010
à transport-...@googlegroups.com
Nope. They just jump a lot higher. Almost every field around here is fenced. Unfortunately, fences need to be 7 or 8 feet tall to contain deer. That gets real expensive real quick.

Jerry Roane

non lue,
19 nov. 2010, 13:51:4019/11/2010
à transport-...@googlegroups.com
Michael

8 foot is not high enough unless the top strand is barbed.  7 foot is a waste of steel.  My pecan orchard has an 8 foot high critter fence and I chased a deer out one day and he jumped the 8 foot fence by rolling over like an Olympic high jumper.  It was amazing to watch this creature belly-roll over my expensive fence to escape me.  I had left the gate open that morning and he got in.  I chased him around on foot to see what he would do and he made a fool out of me with my 8 foot tall fence.  I still need to add the barbed wire top strand.  This was not even a particularly large deer just average sized for the hill country.  They easily jump cattle guards.  When they try an evasive move against cars that is when the danger to everyone happens.  They think they are being tricky moving erratically but they fool the driver and end up road kill.  Also their herding instinct causes a following deer to try to get with the lead deer when they see a car coming that will cut them off.  Since my neighborhood has about 20 permanent deer I see this behavior every time I go to the grocery store around dusk.  

Jerry Roane 
Répondre à tous
Répondre à l'auteur
Transférer
0 nouveau message