T. C. Goldsmith
The potential benefits of automating the guidance of automobiles are extensive especially with regard to better utilization of highway space and safety. Proposals for automobile automation have been made for at least fifty years but a practical system has not been possible because of technology limitations. Now, new advances in technology have brought a practical system within reach. This article discusses the potential benefits of automation, the associated technology requirements, and cost/benefit trades.
To many people the subject of self-guided "automatic" automobiles has a "science fiction" flavor typical of projects that are either far beyond the state of the art or impractical from a cost/benefit standpoint. Actually, recent advances in computers, sensors and other related technology have made such a system feasible in the relatively near term and enormous benefits can justify major development and deployment costs.
Advanced Vehicle Control Systems (AVCS or AVEC) is part of the "Smart Highway" initiative (also known as Intelligent Vehicle Highway Systems (IVHS) or Intelligent Transportation Systems (ITS) now receiving considerable study worldwide.
If being able to take a snooze on the way to Schenectady were the only advantage of an automated car guidance system it would be unlikely that the very substantial development and deployment costs for such a system would be justified in the relatively near future. An automated system can have major advantages over the current system in the areas of highway space utilization and safety as described below.
Human drivers are extremely inefficient in their use of highway space. A typical automobile, when parked in a garage, occupies about 100 square feet of space. Adding "overhead" in the form of areas to open the doors and walk around the car brings the total to perhaps 175 square feet. Yet this same automobile, when operated on the highway at 70 miles per hour requires over 5000 square feet of space. Each commuter, from the time he gets on the highway until he gets off requires an average highway space exceeding one-eighth of an acre that "dynamically" moves with him as he travels in order to operate at 70 mph. This is a large amount of space compared to the space most people occupy to live and work. Depending on the length of "rush-hour" and the length of the average commute only a few other people can reuse the space during peak traffic periods. At 70 mph, each car requires an average of about 250 feet of longitudinal space in a highway lane 12 feet wide in addition to a pro rata share of median strips, clover leafs, and breakdown lanes. Few communities can afford the real estate and construction costs associated with providing this much highway space so, as the density of vehicles increases, traffic tends to slow down until eventually bumper-to-bumper conditions are reached.
Traffic engineers consider that one lane of an optimum highway can carry a maximum of about 2000 cars per hour. Capacity varies with speed from about 750 cars per hour at 5 mph (bumper to bumper) to about 1000 cars per hour at 70 mph. The maximum capacity occurs at 25 35 mph. The cost and other social impact of increasing highway space in or around American cities is easy to imagine.
Why is such a large amount of space required? One major factor is driver reaction time. If one were to write an equation for determining headway (the space between cars on the highway), reaction time would be a major term. Average reaction time for human drivers is probably on the order of two seconds. An automated system could have dramatically reduced reaction time and headway. Another factor is the precision of human drivers. Notice that while cars are about 6 feet wide, highway lanes are 12 feet wide. An automated system could be more precise and therefore require less lateral space.
Although the dynamics of traffic flow, and the characteristics of tires, engines, and steering equipment do require space to operate, by far the largest requirements for highway space are caused by the characteristics of the human vehicle drivers. An automated system could have much faster reaction time and also other characteristics which would dramatically reduce space requirements. The author has estimated that an initial automated system could have space utilization 2.5 times better than the existing system or that two lanes could carry the same traffic as five lanes carry today. This estimate assumes existing vehicles and highways modified by addition of automation. Eventually about 5 times better utilization could be achieved using vehicles and highways specifically designed for automation.
In addition to the space advantage it is reasonable to believe that automated guidance systems could safely operate at top speeds substantially higher than the 70 mph typical in the U.S. today.
The existing automobile/highway system is extremely mature technology having been under continuous development for 100 years. As such we would expect safety advances to be governed by the limitations of "diminishing return".
If we plot accidents, injuries, or deaths per vehicle mile as a function of time (data are available since 1935) we would expect to see incidence decreasing exponentially and approaching a fixed value as the automobiles and highways approach "perfect" and we approach a condition where all accidents are caused by drivers and not other system faults. The data from Virginia is a reasonable approximation of this model. The consequences of simply extending this curve indefinitely to the right forty involve a staggering toll of deaths and injuries. Automobile accidents are now the leading cause of death in certain segments of the population. Currently, approximately 45,000 people die and 1,000,000 are injured in the US. annually. These rates have historically been approximately constant as improvements in the system are offset by increasing traffic. Future medical advances in the cure of diseases can be expected increase the relative future impact of accidents on public health.
To the extent that we could replace safety related driver functions with technology, an automated system could eventually be very substantially safer than the existing system in that we could bring technology to bear directly on a problem that is now virtually completely driver controlled. Vehicle automation could therefore easily be the greatest public health advance of the twenty-first century.
A vehicle guidance system capable of delivering on the promises outlined above would necessarily have to be highly sophisticated and presumably involve substantial electronics, computers, and software. But, vehicle guidance is a very safety critical function. We certainly arent going to deploy a new system that we couldnt prove is safer than the existing system. At the same time, cost is going to be a major factor. Is our technology up to this task?
To explore this issue, lets examine some other transportation systems.
The elevator was first automated in approximately 1940. Because elevators are mechanically guided except for one degree of freedom and other simplifying circumstances, automation could be accomplished without electronics, much less computers. Train automation is somewhat more difficult but also involves mechanical guidance. Mechanically guided automobile systems have actually been proposed but would be much too limited.
Guidance of the Wright brothers flyer (1917) was by means of cables connecting the pilots hands and feet to the control surfaces. Modern aircraft such as the Boeing 747 are guided in the same manner using cables and pulleys with the addition of mechanical/hydraulic force amplification to allow the pilot to control the much larger control surfaces. However, in the 1980s, aircraft such as the Boeing 767 were introduced where guidance is provided by a digital computer system. In effect the computer and associated software controls the plane and the pilots provide advice and direction to the computer via their controls. The computer systems significantly improve safety by detecting and overriding some types of pilot error. These computer systems (which have substantial redundancy) are considered sufficiently reliable to be used as the only means of guiding an aircraft carrying several hundred people and had enough advantages to justify their development and safety certification cost. Fighter aircraft and the Space Shuttle have similar control systems.
Computer capability per dollar has historically improved extremely rapidly. It is therefore not unreasonable to believe that computer systems and associated components with adequate reliability for vehicle guidance will be achievable for "reasonable" cost in the relatively near term.
Keep in mind that the potential benefits of AVCS are extremely large. The savings in highway construction cost, real estate required for highways, pollution, travel time, and reduced injury and death will justify rather large development and deployment costs. Would you rather have a manual Mercedes or an automated Chevrolet that would get you to work in half the time with half the hassle?
One possible approach to vehicle guidance automation would be to simply replace the driver with a "robot" system that would perform some of the same functions, only better, using existing highways. Daimler-Benz has conducted studies of a system which uses computer analysis of TV highway imagery for vehicle guidance. However, a "hybrid" system in which some functions are performed by automation equipment in the highway and other functions are performed by equipment in the vehicle has major advantages and is virtually certain to be chosen for any deployed system. It is assumed that the highway and vehicle systems would communicate and cooperate in the execution of the guidance task. Here are some scenarios illustrating potential features of a hybrid system:
Highways automation systems could have "machine readable" signs, marks, or electronic signals to aid in guidance and supplement any imagery analysis system. The author believes that a second, independent, "backup" method of determining relative vehicle position will be necessary in addition to imagery analysis in order to achieve adequate safety.
Highways could centrally control traffic. Since the highway computer network would have access to sensor data from each vehicle as well as its own sensors it would "know" about conditions everywhere on the highway. It could therefore direct traffic to most efficiently use highway space on a regional basis as well as a lane-by-lane basis. For example, visualize two lanes of traffic where one lane is blocked by construction. In the existing system, each pair of drivers approaching the choke point have to essentially negotiate which is going to go first we all know what that looks like in practice. In a centrally controlled system cars would be directed to interleave well before the choke point. Sight distance would not be an issue. Lateral space could be assigned by the central system based on the size and dynamics of the vehicle as opposed to fixed lanes.
There is some concern that to the extent responsibility for vehicle guidance is transferred to vehicle or highway systems liability exposure will increase for vehicle manufacturers or highway providers. This situation does not appear to be substantively different from the aircraft guidance systems and air traffic control system case and should not be a show stopper.
It is obvious that the successful development and deployment of a hybrid automated vehicle system will require a major commitment on the part of the U.S. federal government to develop and approve the necessary standards for vehicle and highway equipment systems, vehicle and highway certification procedures, vehicle to highway interfaces, etc. As a minimum the governments role would be similar to its existing role in regulation of aircraft systems, interfaces between aircraft and ground systems, etc.
An automated vehicle control system would involve extensive technology in areas where the U.S. has a lead such as computers, networks, software, and aircraft-type systems. If the U.S. took a lead role in developing vehicle automation technology and associated standards, procedures, techniques, and regulations it could be expected to lead the world in vehicle automation in a manner similar to the experience with aviation.
A hybrid automated automobile system as described here would involve system "knowledge" of vehicle movement details and therefore could have privacy issues. However, the same issues apply to credit cards, cell phones, and other modern technology and have not apparently inhibited deployment or use of these systems.
It is unlikely that any near term automated vehicle system will be a "hands-off", "take a nap on the way" system. Passenger aircraft still have pilots. Subways still have drivers. Near term guidance systems will aid and support drivers in a manner similar to the existing aircraft systems. Indeed, one of the most interesting and complex questions to be answered is how to design the interfaces between the automation systems and the drivers for best human engineering.
Early automation systems will probably only accomplish part of the guidance solution. For example, one proposed system would control only the longitudinal "headway" distance to the next vehicle, an advanced type of "cruise control". Such a relatively simple system could achieve increased vehicle density and highway capacity while relieving drivers of considerable traffic stress. Another incremental system possibility would measure a candidate parking place, and then, if it was big enough, automatically park the car, eliminating one of the more unpleasant (but relatively simple to automate) driving chores.
If you enjoyed this article or have comments, criticisms, or questions contact the author at firstname.lastname@example.org.