Professor Berthold Horn
Professor Berthold Horn of Massachusetts Institute of Technology believes a modified adaptive cruise control could mitigate phantom traffic jamsthat occur for no apparent reason.
The phenomenon of the phantom traffic jam is all too common: they appear for no apparent reason and, having caused frustrating delays for all travelers, evaporate for an equally mystical reason. Phantom traffic jams usually occur on busy highways and often take the form of repeatedly stopping and then accelerating up to near the speed limit before coming to a stop again. Professor Berthold Horn of2024 Massachusetts Institute of Technology’s Department of Electrical Engineering and Computer Science has made a study of these traffic jams which are technically known as ‘traffic flow instabilities’. Viewed from overhead, they appear as waves of density and speed that move in the opposite direction to the traffic flow which the Professor denotes as ‘upstream’. These almost periodic instabilities usually take the form of rapid variations in speed and traffic density and create an increased potential for accidents, compromise fuel economy and aaccelerate wear on both vehicles and the road. This is in addition to wasting the drivers’ time and reducing traffic throughput.
While these instabilities have been studied and even modeled by academics, nobody had ever come up with a way to prevent them - until now. Professor Horn has developed a system which he believes could be integrated into adaptive cruise controls to suppress such instabilities.
His solution takes the form of a new algorithm which he presented at the6781 IEEE Conference on Intelligent Transport Systems. Where his approach differs from previous thinking is that he not only takes account of the vehicle ahead but also the one behind.
In general, drivers aim to stay a safe distance behind the vehicle ahead so that if the leading driver were to suddenly apply the brakes, the following driver could stop in time to avoid a collision. When traffic exceeds a certain density, a driver applying the brakes can cause the following driver to apply brakes harder and so on. The Professor found that this wave travels backward, its amplitude increases with distance from the original disturbance and may ultimately cause the traffic at a distant point to come to a complete standstill.
According to the professor this is the vehicular equivalent of a shockwave and comes about because each vehicle only influences what is happening behind. In a very short time, the driver and vehicle performing the initial braking have moved on, meaning the root cause of the congestion has disappeared.
Providing all vehicles can decelerate at the same rate, the research showed that the reaction time of the driver becomes the critical factor and for any given reaction time, there is a corresponding density level above which instabilities will be amplified. While technology could be used to reduce response times and allow vehicles to travel closer to each other at high speed without danger of collision, mathematical modelling indicated that there will still be a critical density above which instabilities occur.
On a practical level the Professor feels there would be potential problems in mixing ‘instrumented’ vehicles with those directly controlled by the driver to whom the automated vehicles will appear to be ‘tailgating’.
Given that a collision is avoided in the braking phase, the vehicles accelerate again and the cycle may repeat itself. To damp out this cyclical and increasingly extreme behaviour, Professor Horn advocates a system which controls the acceleration of the instrumented car in such a way that it remains equidistant between the vehicle ahead and the one behind. He says this will have the effect of dissipating the instability both upstream and downstream.
He likens the effect to having a spring and damper arrangement between the vehicles. If one vehicle in the line applies the brakes, the gap to the following vehicle will become smaller but only briefly and this narrowing will be even smaller in the one behind that. Thus over successive vehicles, the instability will be dissipated and all vehicles will end up travelling at a speed equivalent to the current average in that lane.
The Professor points out that drivers could just start paying more attention to the car behind and not follow the car ahead as closely when there is a bigger gap to the following car. A simple version of this notion is that each driver aims to be halfway between the car behind and the car ahead (think of one spring connecting the car to the one ahead and another to the one behind).
However, for this to happen without automation a high proportion of drivers would have to adopt this technique if the instability is to be dissipated. This is both undesirable, because paying attention to the car behind distracts the driver’s attention from the one in front which could reduce safety, and unlikely as some drivers are likely to loathe leaving gaps ahead which vehicles from other lanes could occupy. The Professor’s preference is for a bilateral adaptive cruise control system with sensors to determine the relative positions and speeds of the vehicles ahead and behind and to then automatically control the instrumented vehicle’s acceleration.
So whereas a normal cruise control maintains a fixed speed, the system proposed by the Professor follows the bilateral control law and attempts to match the average speed of the leading and the following vehicles while staying more or less equidistant from both. Being automated the system would require no driver input and would therefore not compromise safety while additional control rules would take into account minimum safe separation, relative speeds, speed limits, weather and lighting conditions, traffic density and traffic advisories. If there are no nearby vehicles, the system could revert to a simple cruise control mode where the acceleration is simply a function of the current speed and the desired speed.
According to the Professor, the sensors required to implement this type of control system would be in the vehicle - such as cameras aimed fore and aft using machine vision methods to estimate the velocity of and distance to the nearest vehicle. Laser, ultrasound and radar measurement technique could be used, although he feels they are likely to be more expensive and less able to discriminate between vehicles in the same lane and those in adjacent lanes.
He describes the effect thus: “Suppose that a number of vehicles using this new type of control system start off with different initial velocities and different inter-vehicle spacing. The effect of the ‘damper’ part of the control system is to dissipate the kinetic and potential energy resulting from departures from the average [speed and spacing], thus reducing the difference between individual vehicle’s motions and the average.”
This raises the potential for a number of similarly instrumented vehicles to form a type of platoon but one in which there is no direct communication between the vehicles and no single driver is in control of the group. The advantage over the standard platoon model is that it would be easy for cars to drop out of the group and other cars to join. Also, the group of cars can easily react to disturbances and would then automatically revert towards an equilibrium phase.
It may also be easier and more acceptable to mix instrumented vehicles travelling in this manner with ‘normal’ non-instrumented vehicles than it would be for a conventionally configured platoon.
In concluding the Professor said: “If most cars had this kind of system, then traffic would flow more smoothly at high densities, and existing roadways could sustain higher throughputs. Gas consumption would be reduced, as will time lost sitting in stop-and-go traffic. Wear and tear on roads and vehicles would be lessened and the incidence of accidents reduced. Community benefits could be significant, since more traffic could be accommodated without adding to the infrastructure.”
This article was produced with the permission of the Massachusetts Institute of Technology and the full paper can be downloaded from <%$Linker:2 External <?xml version="1.0" encoding="utf-16"?><dictionary /> 0 0 0 oLinkExternal here Visit people.csail.mit.edu/bkph/articles/Suppressing_Traffic_Flow%20Instabilities_IEEE_ITS_2013.pdf false http://people.csail.mit.edu/bkph/articles/Suppressing_Traffic_Flow%20Instabilities_IEEE_ITS_2013.pdf false false %>
The phenomenon of the phantom traffic jam is all too common: they appear for no apparent reason and, having caused frustrating delays for all travelers, evaporate for an equally mystical reason. Phantom traffic jams usually occur on busy highways and often take the form of repeatedly stopping and then accelerating up to near the speed limit before coming to a stop again. Professor Berthold Horn of
While these instabilities have been studied and even modeled by academics, nobody had ever come up with a way to prevent them - until now. Professor Horn has developed a system which he believes could be integrated into adaptive cruise controls to suppress such instabilities.
His solution takes the form of a new algorithm which he presented at the
In general, drivers aim to stay a safe distance behind the vehicle ahead so that if the leading driver were to suddenly apply the brakes, the following driver could stop in time to avoid a collision. When traffic exceeds a certain density, a driver applying the brakes can cause the following driver to apply brakes harder and so on. The Professor found that this wave travels backward, its amplitude increases with distance from the original disturbance and may ultimately cause the traffic at a distant point to come to a complete standstill.
According to the professor this is the vehicular equivalent of a shockwave and comes about because each vehicle only influences what is happening behind. In a very short time, the driver and vehicle performing the initial braking have moved on, meaning the root cause of the congestion has disappeared.
Providing all vehicles can decelerate at the same rate, the research showed that the reaction time of the driver becomes the critical factor and for any given reaction time, there is a corresponding density level above which instabilities will be amplified. While technology could be used to reduce response times and allow vehicles to travel closer to each other at high speed without danger of collision, mathematical modelling indicated that there will still be a critical density above which instabilities occur.
On a practical level the Professor feels there would be potential problems in mixing ‘instrumented’ vehicles with those directly controlled by the driver to whom the automated vehicles will appear to be ‘tailgating’.
Given that a collision is avoided in the braking phase, the vehicles accelerate again and the cycle may repeat itself. To damp out this cyclical and increasingly extreme behaviour, Professor Horn advocates a system which controls the acceleration of the instrumented car in such a way that it remains equidistant between the vehicle ahead and the one behind. He says this will have the effect of dissipating the instability both upstream and downstream.
He likens the effect to having a spring and damper arrangement between the vehicles. If one vehicle in the line applies the brakes, the gap to the following vehicle will become smaller but only briefly and this narrowing will be even smaller in the one behind that. Thus over successive vehicles, the instability will be dissipated and all vehicles will end up travelling at a speed equivalent to the current average in that lane.
The Professor points out that drivers could just start paying more attention to the car behind and not follow the car ahead as closely when there is a bigger gap to the following car. A simple version of this notion is that each driver aims to be halfway between the car behind and the car ahead (think of one spring connecting the car to the one ahead and another to the one behind).
However, for this to happen without automation a high proportion of drivers would have to adopt this technique if the instability is to be dissipated. This is both undesirable, because paying attention to the car behind distracts the driver’s attention from the one in front which could reduce safety, and unlikely as some drivers are likely to loathe leaving gaps ahead which vehicles from other lanes could occupy. The Professor’s preference is for a bilateral adaptive cruise control system with sensors to determine the relative positions and speeds of the vehicles ahead and behind and to then automatically control the instrumented vehicle’s acceleration.
So whereas a normal cruise control maintains a fixed speed, the system proposed by the Professor follows the bilateral control law and attempts to match the average speed of the leading and the following vehicles while staying more or less equidistant from both. Being automated the system would require no driver input and would therefore not compromise safety while additional control rules would take into account minimum safe separation, relative speeds, speed limits, weather and lighting conditions, traffic density and traffic advisories. If there are no nearby vehicles, the system could revert to a simple cruise control mode where the acceleration is simply a function of the current speed and the desired speed.
According to the Professor, the sensors required to implement this type of control system would be in the vehicle - such as cameras aimed fore and aft using machine vision methods to estimate the velocity of and distance to the nearest vehicle. Laser, ultrasound and radar measurement technique could be used, although he feels they are likely to be more expensive and less able to discriminate between vehicles in the same lane and those in adjacent lanes.
He describes the effect thus: “Suppose that a number of vehicles using this new type of control system start off with different initial velocities and different inter-vehicle spacing. The effect of the ‘damper’ part of the control system is to dissipate the kinetic and potential energy resulting from departures from the average [speed and spacing], thus reducing the difference between individual vehicle’s motions and the average.”
This raises the potential for a number of similarly instrumented vehicles to form a type of platoon but one in which there is no direct communication between the vehicles and no single driver is in control of the group. The advantage over the standard platoon model is that it would be easy for cars to drop out of the group and other cars to join. Also, the group of cars can easily react to disturbances and would then automatically revert towards an equilibrium phase.
It may also be easier and more acceptable to mix instrumented vehicles travelling in this manner with ‘normal’ non-instrumented vehicles than it would be for a conventionally configured platoon.
In concluding the Professor said: “If most cars had this kind of system, then traffic would flow more smoothly at high densities, and existing roadways could sustain higher throughputs. Gas consumption would be reduced, as will time lost sitting in stop-and-go traffic. Wear and tear on roads and vehicles would be lessened and the incidence of accidents reduced. Community benefits could be significant, since more traffic could be accommodated without adding to the infrastructure.”
This article was produced with the permission of the Massachusetts Institute of Technology and the full paper can be downloaded from <%$Linker:
