Self-Drive
Cars: What’s the point?
Let me state
from the outset, that I have no training or knowledge about the technology
behind self-drive cars. I haven’t taken an interest in them, and have never had
a strong opinion about them. After all, what’s the point? It’s a dumb idea that
will never work, right?
To be fair,
that probably is a strong opinion; so let’s try to understand the idea before
we discard it.
People at
Google, and other “Brainiac Factories” are trying to build a car that will do
the driving for us. Basically that must involve building into said car, a
myriad of sensors that can “see” the environment it is driving in, and react in
the same way that a human driver would. No, not the same way. Better.
All my life, I
have been hearing about the ambitions of computer programmers to create a chess
playing opponent that can beat a grand master. Have they achieved it? Don’t
know. The world lost interest. I know I did, and I quite like both computers
and chess.
Then there were
the boffins trying to create artificial intelligence. Apparently, the test is
for somebody to ask it a series of questions, and for it to give structured,
“considered” replies. The questions would be structured in such a way that the
answers could not be scripted, and would demonstrate actual thought. Have they
achieved it? Not that I have heard.
Given the less
than stellar progress in replicating specific, focused aspects of the human
brain, it seems absurd that somebody would think that it’s a cool idea to ramp
up the difficulty factor. They don’t want to build a car that will drive on a
test track without incident. They want it to be able to drive in traffic, on
unfamiliar roads, and share that road with the most unpredictable and dangerous
animal on the planet. Me.
No, I was right.
It’s a dumb idea.
But…
There is merit
in a controlled environment where ALL cars are controlled, and ALL cars can
communicate with each other for a common result. A “Hive” mentality.
So how about
this:
·
A car that I can drive when I
need to.
·
A car that “slots in” when in a
high volume environment. Like a freeway?
So if we think
about the slotcar model for a jumping off point. The driver can control many
things, but when it’s in the slot, “the road” controls the steering.
Can we create
freeways with slotted grooves down the centre of each lane, for cars to drop a
“guide flag” into? Probably not.
But we can lay
metallic strips that the car can magnetically “sync” with, to stay on track.
So I drive my
car from home to the freeway on-ramp:
·
As I line up, I drive across
the guiding strips on the ramp, and the car is guided into position.
·
At this point, I relinquish
steering control o/f the vehicle. Just like some of those sci-fi movies, the
steering wheel might retract into the dashboard.
·
I also relinquish speed
control, so the pedals could retract as well.
·
Using the “hive” analogy, the
freeway then absorbs my car into its neural network. We are slotted in between
other cars going the same direction. Our GPS would have told the network of our
destination, so it will guide us to the correct exit. In the meantime, we
travel at optimal speed, with all the other worker bees.
·
The Neural Network will know
how busy the traffic is at any moment in time, so will use all available lane
in the most efficient way possible. For some stretches of freeway, there may be
8 lanes moving north, with 2 moving south. Allocation of direction for each
lane will be dynamically recalculated constantly, and the speed of vehicles in
corresponding zones adjusted accordingly.
·
The network may decide that
certain lanes will move faster than others. When a car on the faster inner lane
was approaching its exit or freeway change, the network will adjust the speed
of vehicles nearby, to facilitate it moving seamlessly into the outer lanes in
time to make the exit.
I see this
concept as very exciting. Imagine being able to stop the gridlock on freeways
and motorways around the world. I also see it as far more practical and
achievable than the overly ambitious and hazardous path currently being explored.
After all,
while this idea does utilise self-drive, it does not need all of the
human-emulating traits that these people are trying to create. It is
controlling a vehicle, on a confined freeway, with no aberrations or human
anomalies to contend with, and it is magnetically locking the vehicle to a
“track” that follows a predefined path.
When the
vehicle needs to exit the freeway, it switches between magnetic lanes until it
is following the path up the exit ramp. I imagine these magnetic strips are not
just one-per-vehicle lane. I imagine that there will be a bank of strips, only
centimetres apart, and the magnetic lock under each car would be reading across
many strips at once. In this way, the car could be instructed to move left or
right in small or large increments, while staying “in contact” with the road at
all times.
If the network
was aware of the physical dimensions of each vehicle, then it could use this
incremental left/right adjustment to stack vehicles tightly across the width of
the freeway. What may be 4 lanes when the big trucks are moving, could be 6
lanes when the soccer mums are out in force.
Because the
network also controls the speed of all vehicles, they can travel both faster
and closer than would be “humanly” possible. Lateral and longitudinal
separation can be reduced to millimetres at quite high speeds, when a lane
change is required.
If a vehicle
detects a fault requiring immediate attention, the same method is used to
“herd” the vehicle to the side of the road, where it can stop safely.
Uptake
Laying network strips on the surface of freeways
is probably not as prohibitively expensive as some alternatives. The important
thing is that only compatible cars be allowed onto the freeway. I would imagine
that if non-compatible cars were forced onto second-tier roads for the same
journey, then there would be considerable incentive to buy a compatible model,
o



