The cybernetic DWIM car is coming. DWIM stands for "Do What I
Mean."1 It is a
commonplace term in the field of Human-machine Interfaces, and refers to systems
that automatically interpret the user's intent from his or her inputs.
Cybernetics (or at least one aspect of it) is the science of unifying humans
and machines. The objective of cybernetics is usually to amplify human
capability with "intelligent" machines, but sometimes the objective is
the reverse. Most of the work in cybernetics has been under the aegis of
defence, for building advanced tanks and aircraft. There is a modest amount of
cybernetics in the automotive industry, as well. Anti-lock Braking (ABS),
Acceleration Slip Reduction (ASR), Electronic Engine Management, and Automatic
Traction Control (ATC) are cybernetic DWIM systems---of a kind---already in
production. They all make "corrections" on the driver's input based on
an assumed intention. Steer-by-wire, Continuously Variable Transmissions (CVT),
and active suspensions are on the immediate horizon. All these features are part
of a distinct trend to automate the driving experience. This month, we take a
break from hard physics to look at the better and the worse of increased
automation, and we look at one concept of the ultimate result, CyberCar.
Among the research directions in cybernetics are advanced sensors for human
inputs. One of the more incredible is a system that reads brain waves and
figures out what a fighter pilot wants to do directly from patterns in the
waves.
A major challenge in the fighter cockpit is information overload. Pilots have
far too many instruments, displays, horns, buzzers, radio channels, and idiot
lights competing for their attention. In stressful situations, such as high
speed dogfights, the pilot's brain simply ignores inputs beyond its capacity, so
the pilot may not hear a critical buzzer or see a critical warning light. In the
"intelligent cockpit," however, the pilot consciously
suppresses certain displays and auditory channels, thus reducing sensory
clutter. By the same token, the intelligent cockpit must be able to override the
pilot's choices and to put up critical displays and to sound alarms in
emergencies. In the reduced clutter of the cockpit, then, it is much less likely
that a pilot will miss critical information.
How does the pilot select the displays that he2 wants to see? The pilot cannot afford the
time to scroll through menus like those on a personal computer screen or
hunt-and-peck on a button panel like that on an automatic bank teller machine.
There are already sensors that can read a pilot's brain waves and anticipate
what he wants to look at next. Before the pilot even consciously knows that he
wants to look at a weapon status display, for example, the cybernetic system can
infer the intention from his brain waves and pop up the display. If he thinks it
is time to look at the radar, before he could speak the command, the system
reads his brain waves, pops up the radar display, and puts away the weapon
status display.
How does it work? During a training phase, the system reads brain waves and
gets explicit commands through a button panel. The system analyses the brain
waves, looking for certain unique features that it can associate with the
intention (inferred from the command from the button panel) to see the radar
display, and other unique features to associate with the intention to look at
weapon status, and so on. The system must be trained individually for each
pilot. Later, during operation, whenever the system sees the unique brain wave
patterns, it "knows" what the pilot wants to do.
The implications of technology like this for automobiles is amazing. Already,
things like ABS are a kind of rudimentary cybernetics. When a driver stands all
over the brake pedal, it is assumed that his intention is to stop, not to skid.
The ABS system "knows," in a manner of speaking, the driver's
intention and manages the physical system of the car to accomplish that goal.
So, instead of being a mere mechanical linkage between your foot and the brakes,
the brake pedal becomes a kind of intentional, DWIM control. Same goes for
traction control and ASR. When the driver is on the gas, the system
"knows" that he wants to go forward, not to spin out or do doughnuts.
In the case of TC, the system regulates the torque split to the drive wheels,
whether there be two or four. In the case of ASR, the system backs off the
throttle when there is wheel spin. Cybernetics again.
ABS, TC, and ASR exist now. What about the future? Consider steer-by-wire.
CyberCar, the total cybernetic car, infers the driver's intended direction from
the steering wheel position. It makes corrections to the actual direction of the
steered wheels and to the throttle and brakes much more quickly and smoothly
than any driver can do. Coupled with slip angle3 sensors [1]
and inertial guidance systems, perhaps based on miniaturized laser/fibre optic
gyros (no moving parts), cybernetic steering, throttle, and brake controls will
make up a formidable racing car that could drive a course in practically optimal
fashion given only the driver's desired racing line.
In an understeering situation, when a car is not turning as much as desired,
a common driver mistake is to turn the steering wheel more. That is a mistake,
however, only because the driver is treating the steering wheel as an intentional
control rather than the physical control it actually is. In CyberCar, however,
the steering wheel is an intentional control. When the driver adds more
lock in a corner, CyberCar "knows" that the driver just wants more
steering. Near the limits of adhesion, CyberCar knows that the appropriate physical
reaction is, in fact, some weight transfer to the front, either by trailing
throttle or a little braking, and a little less steering wheel lock. When the
fronts hook up again, CyberCar can immediately get back into the throttle and
add a little more steering lock, all the while tracking the driver's desires
through the intentional steering wheel in the cockpit. Similarly, in an
oversteer situation, when the driver gives opposite steering lock, CyberCar
knows what to do. First, CyberCar determines whether the condition is trailing
throttle oversteer (TTO) or power oversteer (PO). It can do this by monitoring
tyre loads through suspension deflection and engine torque output over time. In
TTO, CyberCar adds a little throttle and counter steers. When the drive wheels
hook up again, it modulates the throttle and dials in a little forward lock. In
PO, CyberCar gently trails off the throttle and counter steers. All the while,
CyberCar monitors driver's intentional inputs and the physical status of the car
at the rate of several kilohertz (thousands of times per second).
The very terms "understeer" and "oversteer" carry
cybernetic implication, for these are terms of intent. Understeer means the car
is not steering as much as wanted, and oversteer means it is steering too much.
The above description is within current technology. What if we get really
fantastic? How about doing away with the steering wheel altogether? CyberCar,
version II, knows where the driver wants to go by watching his eyes, and it
knows whether to accelerate or brake by watching brain waves. With Virtual
Reality and teleoperation, the driver does not even have to be inside the car.
The driver, wearing binocular video displays that control in-car cameras (or
even synthetic computer graphics) via head position, sits in a virtual
cockpit in the pits.
Now we must ask how much cybernetics is desirable? Autocrossing is, largely,
a pure driver skill contest. Wheel-to-wheel racing adds race craft--drafting,
passing, deception, etc. --to car control skills. Does it not seem that
cybernetics eliminates driver skill as a factor by automating it? Is it not just
another way for the "haves" to beat the "have-nots" by
out-spending them? Drivers who do not have ABS have already complained that it
gives their competition an unfair advantage. On the other hand, drivers who do
have it have complained that it reduces their feel of control and their options
while braking. I think they doth protest too much.
In the highest forms of racing, where money is literally no object,
cybernetics is already playing a critical role. The clutch-less seven speed
transmissions of the Williams/Renault team dominated the latter half of the 1991
Formula 1 season. But for some unattributable bad luck, they would have won the
driver's championship and the constructor's cup. Carrol Smith, noted racing
engineer, has been predicting for years that ABS will show up in Formula 1 as
soon as systems can be made small and light enough [2]. It seems inevitable to me that cybernetic systems will
give the unfair advantage to those teams most awash in money. However,
autocrossers, club racers, and other grass roots competitors will be spared the
expense, and the experience of being relieved of the enjoyment of car control,
for at least another decade or two.
Acknowledgements
Thanks to Phil Ethier for giving me a few tips on car control that I
might be able to teach to CyberCar and to Ginger Clark for bringing slip angle
sensors to my attention.
