00100			COMPUTER CONTROLLED CARS
00150	
00175			by John McCarthy, Stanford University
00200	
00300		There have been a number of proposals for  automatic  control
00400	of  cars.     Mostly, they have involved simple servo-mechanisms that
00500	sense a cable buried in the roadway  and  some  other  mechanism  for
00600	sensing  the distance of the car ahead.  Such a scheme was studied at
00700	RCA at the instigation of  Zworykin,  but  the  work  was  eventually
00800	abandoned.
00900	
01000		In  science  fiction,  systems  in  which  a  single computer
01100	controls all the cars in a wide area have been depicted  but  without
01200	telling how the system would actually work.
01300	
01400		We  are  also proposing the computer control of cars.     Our
01500	system requires a computer in the car equipped with television camera
01600	input  that uses the same visual input available to the human driver.
01700	Essentially, we are proposing an automatic chauffeur.  Our goal is  a
01800	system with the following qualities:
01900	
02000		1.   The user enters the destination with a keyboard, and the
02100	car drives him there.  Other commands include:   change  destination,
02200	stop at that rest room or restaurant, go slow, go at emergency speed.
02300	
02400		2.  The user need not be a driver and need not even accompany
02500	the car.      This would permit children, old people, and  the  blind
02600	greater personal freedom.   It also permits a husband to be driven to
02700	work, then send the car home for his wife's use, and permits  her  to
02800	send it back for him at the end of the day.   The car can be sent for
02900	servicing or to a store where a telephone-ordered  purchase  will  be
03000	put  in  it.     If there is a suitable telephone system, the car can
03100	deliver a user to a place where there is  no  parking,  go  away  and
03200	park,  and  return when summoned.  Thus, the system is to have almost
03300	all of the capabilities of a chauffeur.
03400	
03500		3.   In contrast to a system based on a central computer, the
03600	proposed  system  will  be  of advantage to the first person who buys
03700	one, whether anyone else has it or not. It will require no change  in
03800	existing  roads, but will be able to take orders from traffic control
03900	computers when they are installed.      When  freeway  lanes  can  be
04000	dedicated to computer controlled cars they will multiply the capacity
04100	of existing freeways by permitting 80 mile per hour  bumper-to-bumper
04200	traffic  with greater safety than we have at present.       Since the
04300	system is a product and  not  a  public  utility,  competition  among
04400	suppliers will be possible.
04500	
04600		4.    A key goal is to achieve greater safety than we have at
04700	present.  A five-fold reduction in fatalities is probably required to
04800	make  the  system  acceptable.   Much better is possible since humans
04900	really  are  rather  bad  drivers,  but  complete  safety  cannot  be
05000	guaranteed.
05100	
05200		Now  we shall consider the problems that have to be solved in
05300	order to realize the system.
05400	
05500		1.  Performance of  the  computer,  cameras,  and  associated
05600	electronics.
05700		Present computers seem to be fast enough and to  have  enough
05800	memory  for  the job.   However, commercial computers of the required
05900	performance are too big.  We envisage that a computer  of  about  the
06000	power  of  the  Digital Equipment PDP-10 will be required.   Military
06100	versions of similar computers have volumes of one or two cubic  feet,
06200	but  the  requirements  for  memory  and  secondary  storage would be
06300	difficult to meet in a reasonable volume at present.    However,  the
06400	development  of more compact computers and other electronic circuitry
06500	is proceeding at a rate that makes achieving the required compactness
06600	not  the  pacing  item.      Some improvements in the performance and
06700	compactness of television cameras is also required, but it is not yet
06800	clear what these requirements are.
06900	
07000		2. Cost of the computer and other electronics.
07100		At present prices, a computer capable of  controlling  a  car
07200	but  containable only in a large van would cost $400,000 to $800,000.
07300	A few thousand dollars worth of other electronics would be  required.
07400	Ten  years  should  bring  the  cost down by a factor of ten.    Mass
07500	production would give another factor of three.  This would permit the
07600	system  to  be available as a luxury item.  Another five to ten years
07700	might be required before computer control would only double the price
07800	of the car.  These estimates must be regarded as guesses.
07900	
08000		3. Reliability of the computer and other electronics.
08100		We  can  attempt  to  compute  the  required  reliability  by
08200	demanding  that  present traffic fatalities be reduced to a fifth the
08300	present number, i.e.  to 10,000 per year, and by allocating only half
08400	of  these  fatalities  to unreliability of the electronics.      This
08500	further depends on the fraction of failures  that  lead  to  fatality
08600	which  can  be kept quite low by having the computer check its health
08700	and that of the electronics  every  tenth  of  a  second,  giving  it
08800	programs for dealing with partial failures, and providing a "dead man
08900	switch" for stopping the car if the computer  fails  to  reassure  it
09000	every  tenth  second.  There are many possibilities in this direction
09100	and the expenditure of much cleverness is called for.  The reader  is
09200	advised  against using his unaided intuition to estimate the results.
09300	Nevertheless, present computer failure rates would not be  acceptable
09400	even   if   they   never  led  to  accident  simply  because  of  the
09500	inconvenience.      We  estimate  that  an  improvement  of  1000  in
09600	mean-time-between-failures  is  required.     Rapid progress is being
09700	made in this field, and we expect that ten to  fifteen  years  normal
09800	progress of the computer field will give the required result.
09900	
10000		4.   Performance of the driving programs.
10100		Developing  the  required  computer  programs  is  the   most
10200	difficult  of  the  required tasks; it will probably take the longest
10300	time; and the amount of time required is very difficult  to  predict.
10400	Work  on  computer  control  of  vehicles has started at the Stanford
10500	University Artificial Intelligence Project. An  experimental  vehicle
10600	has  been  equipped  with  a  television  camera and connected to the
10700	computer with a two-way TV and radio link.     A  simple  program  to
10800	guide the vehicle to follow a white line like that in a road has been
10900	successfully checked out, and programs for determining the course  of
11000	the  road and detecting cars and other obstacles are being developed.
11100	However, before computer controlled cars  become  a  reality  a  much
11200	larger scale effort will have to be made.
11300	
11400		The  nature  and extent of this effort are not easy to forsee
11500	yet.    We are far from having exhausted  the  possibilities  of  our
11600	present equipment, but eventually the radio link to the computer will
11700	have to be replaced by a computer  in  the  vehicle,  and  television
11800	equipment  capable  of  seeing better into shadows in the presence of
11900	bright areas will be required.  We need to be able to  identify  many
12000	different  types of objects on the road such as:   persons, vehicles,
12100	animals, traffic police, shadows, pieces of paper,  cardboard  boxes,
12200	objects  that  have  dropped  from  vehicles, traffic signs and other
12300	signals, intersections, house numbers, and other information required
12400	for  navigation.    It  will  have  to  be  equipped with programs to
12500	recognize and deal with a variety of emergency conditions.   It  will
12600	surely  be  possible  to make it better at this than humans since its
12700	attention won't lapse, it can sense the mechanical condition  of  the
12800	car  continuously,  and  it can look to the side, underneath the car,
12900	and behind every second.
13000	
13100		The most intricate  single  problem  is  the  visual  pattern
13200	recognition.
13300	
13400		5. Testing.
13500		After the required performance is demonstrated and before the
13600	system  can  be  trusted  without a human driver an extensive testing
13700	program is required.   To demonstrate that the system is  five  times
13800	safer  than a human driver approximately 25,000 automobile years will
13900	be required. This might be reduced somewhat by concentrating  testing
14000	on  situations in which humans make most of their fatal mistakes, but
14100	we would still need to be sure that situations in which  the  program
14200	made fatal blunders peculiar to the computer system were rare enough.
14300	Developments in the mathematical theory  of  computation  may  permit
14400	getting  rid of ordinary programming errors and proving that they are
14500	absent, but possible inadequacies in the  algorithms  themselves  can
14600	only be obviated by testing.
14700	
14800		6.  Public acceptance.
14900		Automobiles without  qualified  human  drivers  will  require
15000	changes  in the law.  Fortunately, testing such systems with a driver
15100	present to take over if necessary does not. Moreover, computer driven
15200	cars will not be able to obey oral instructions from policemen , so a
15300	digital system will have to be developed.  A  general  resistance  to
15400	technological  innovation  on  the  part of the literary culture will
15500	have to be overcome, but it seems to me that after the test phase the
15600	advantages will be clear enough so that this will not be difficult.
15700	
15800		7. Support for research and development.
15900		The  development  of  computer  controlled  cars  will   cost
16000	hundreds  of  milliions  of  dollars.   A computer program capable of
16100	reliably taking care of all  the  contingencies  that  can  arise  in
16200	driving a car will have to be more complex than any ever written, and
16300	adequate testing will require a complex  organization.   Fortunately,
16400	The  commitment of large amounts of money will be required only after
16500	spectacular   though   unreliable   performance   will   have    been
16600	demonstrated.  So far as I know, the Stanford Artificial Intelligence
16700	Project is the only organization now working on computer control of a
16800	vehicle  using vision.  This work is part of a basic research project
16900	on artificial intelligence supported by the  Department  of  Defense.
17000	Even  at  the present stage of the work, other projects are needed to
17100	secure  an  adequate  diversity  of  approach.    While  considerable
17200	additional  progress will certainly be made with the present support,
17300	even a prototype will require more money than is now available.
17400	
17500		Fortunately, this problem is within the jurisdiction  of  the
17600	Department  of  Transportation.    The  automobile  companies and the
17700	computer companies also might be expected to  help,  but  their  past
17800	record  of  seeing beyond the ends of their noses is not encouraging.
17900	Because the programming is the pacing item, more support at this time
18000	will  hasten  the  day when computer control of cars is achieved, but
18100	the possibilities will be much more obvious in five  years  with  the
18200	advances  in  hardware  and programming that are already taking place
18300	for other reasons.
18400	
18500		Finally, we would like to deal with some arguments that might
18600	be  raised  against  supporting research aimed at computer-controlled
18700	cars:
18800	
18900		1.   Cars must be done away with because they  produce  smog,
19000	require too much space, and use up too much natural resources.
19100		We believe the smog devices will eventually be made  to  work
19200	well,  or if not, another form of propulsion can be found.   Computer
19300	controlled cars will require less space than equivalent present  cars
19400	because  they can go faster and closer together on streets, roads and
19500	freeways, because they can park at a distance from a place where they
19600	discharge passengers, and because a computer driven car can be shared
19700	more easily than a conventional car.  If hydrocarbon  fuel  runs  out
19800	and is still required for cars, then with nuclear energy, the burning
19900	reaction can be driven backwards and  fuel  synthesized  from  carbon
20000	dioxide and water.
20100	
20200		2. A simpler scheme of automatic control is preferable.
20300		The buried cable and other simple  schemes  do  not  increase
20400	human  freedom  and  convenience.     They  only permit us to use the
20500	freeways a bit more efficiently.    Because  of  their  inability  to
20600	detect  dogs,  children,  potholes, and objects that have fallen from
20700	trucks they may require unrealizable control of access to the highway
20800	in order to achieve safety.
20900	
21000		3.    Some form of automated mass transportation is obviously
21100	better.
21200		The automobile can go point to point in areas of both low and
21300	high density.   We believe that these advantages should not and  will
21400	not  voluntarily  be  given up.  We favor the development of improved
21500	mass transportation, but predict that the automobile will be given up
21600	only  for  something  that  works  better  in  all  ways  such as an
21700	individual computer controlled flying machine  capable  of  point  to
21800	point transportation.