EP0750774B1 - Method of detecting traffic and traffic situations on roads, preferably motorways - Google Patents

Abstract

PCT No. PCT/DE95/00265 Sec. 371 Date Sep. 12, 1996 Sec. 102(e) Date Sep. 12, 1996 PCT Filed Mar. 1, 1995 PCT Pub. No. WO95/25321 PCT Pub. Date Sep. 21, 1995A method of sensing traffic and detecting traffic situations on roads (AS), preferably freeways. With measuring points (measuring cross sections MQ1, MQ2, . . . ) set up for the purpose for vehicle detection using traffic sensors (VS) and with a traffic data processing arrangement (VDVE) for traffic control, at regular intervals traffic data (VD), such as vehicle speed (v), traffic intensity (Q) and traffic density (K), are determined and traffic parameters determined therefrom are formed in a traffic data processing system (VDA). Two adjacent measuring points. (MQi, MQ(i+1) form a measuring section (MA) of a given length (l). The following traffic parameters are formed from the traffic data (VD) of two such measuring points: a) the speed density difference (vk-D), which is calculated from the local traffic data of average speed (v) and traffic density (K); b) a trend factor (FT), which is formed continually from the ratio between the traffic intensities (Qi/Q(i+1)) of the first and second measuring points (MQi, MQ(i+1)), but determined during a given period (t) in the minute range; c) the traffic intensity trend (QTi, QT(i+1)) of the respective measuring point (MQi, MQ(i+1)), the trend being derived on the basis of the function of the traffic intensity (Q) over the time (curve Q (t)) from the increase of the tangent to the curve. The probability of a critical traffic situation (WG) is derived therefrom in a fuzzy logic (FUB).

Description

The invention relates to a method for traffic detection and traffic situation detection on highways according to the preamble of claim 1.

The ever increasing traffic on the street, preferably Autobahn, with the resulting loss of security and the difficulties, the entire road or highway network have expanded accordingly in recent years Decades led to considerations of performance as well as the safety of the highways with the help of electronics to increase.

In the meantime there are various plants and different ones Procedures that display appropriate ads based on traffic measurements turn on the variable message sign. The control is point based if it is on a certain point of the traffic flow (e.g. construction sites or Road narrowing), due to the route (generally under known as "line manipulation") if they are refers to a route, or network-related if it is the automatic rerouting from a normal route to an alternative route undertakes (alternating directions).

Current line influencing is very complex and expensive and are therefore only targeted at particularly busy routes built. This is a very high effort in terms of Data acquisition and evaluation as well as for information transfer required by means of variable message signs. Around the relationships between the traffic situation and the controlled one The control logic is to keep the display manageable relatively simple. The prepared local measured values, like generally smoothed traffic, smoothed Speed and local traffic density are predefined with Threshold values compared to make a statement or to control the variable message sign.

For systems in operation with traffic detection and Control of traffic through variable message signs is so far control based on clear yes / no statements on decision logic. For example a harmonization of traffic achieved due to high loads be the same on all lanes Speed limit is displayed. A traffic jam detection and warning may be due to a decrease in driving speed respectively. Becomes a relatively heavy traffic unrest recognized, so can with an even speed limit be met. Weather dependent Ambient conditions detected with separate sensors are also displayed to influence the line. Early and reliable detection of dangerous Traffic conditions are not without the known systems further possible because the recorded traffic data none clear information about the actual traffic situation surrender.

EP-A-0 171 098 describes a method for traffic detection and disclosed for traffic control on highways, which at least two measuring points for vehicle detection with traffic sensors having. Traffic data is in the form of vehicle speeds taking into account the traffic volume determined, processed and evaluated. It will there the determined speed data of at least two, considered in a certain length of spaced measuring points and based on logical decisions with predefined ones Speed values compared.

From the article "Traffic Prediction Method by Fuzzy Logic", Second IEEE International Conference on Fuzzy Systems (Cat. No. 93CH3136-9), Proceedings of IEEE 2nd International Fuzzy Systems Conference, San Francisco, CA, USA, March 28 - 1 April 1993, ISBN 0-7803-0614-7, 1993, New York, NY, USA, IEEE, USA, pages 673-678 vol. 2, Iokibe T. et al. is a Method known in which only the traffic volume measured and this together with experience of Traffic levels are assessed using fuzzy logic get an estimate of the expected traffic.

The object of the invention is an early and reliable, automatic detection of critical traffic situations, such as traffic disruptions caused by congestion or an accident, on motorways to the road users in good time before this Warning situation.

This task is carried out in the method described at the beginning solved with the characterizing features of claim 1.

With the method according to the invention on the highways with measuring points set up for this purpose, these are the respective measuring cross sections with traffic sensors installed for each lane, Traffic data recorded and in a processing device provided for this purpose processed for traffic control. From the regularly recorded traffic data: speed and traffic volume, are in a traffic data processing device certain traffic parameters derived. For this purpose, two adjacent measuring points form a measuring section, which has a certain distance, for example 3 km. From the traffic data from these measuring points The following traffic quantities are formed:

A speed density difference (vk-D) according to the relationship as specified in claim 1. The speed density difference takes into account the speed and the traffic density of both measurement cross sections. As a second traffic parameter a trend factor is formed that is continuous from the ratio of the heavy traffic of the first and the second measuring point is formed, but only the values above one certain period, e.g. the last 30 minutes. The third traffic parameter is the traffic volume trend the respective measuring point as a measure of the dynamic Situation development, i.e. the temporal development the traffic volume. The function of Traffic intensity over time or from the slope of the tangent on this functional curve the trend of traffic volume derived. These three traffic parameters are based on fuzzy logic processed to detect critical traffic situations, to make a statement about the probability as an output variable for a critical traffic situation. This probability quantity is dependent on a predeterminable threshold value evaluated to a display recommendation to generate for the variable message signs.

The application of fuzzy logic for traffic situation detection on highway roads has a number of advantages. The evaluation the input data is very simple. Multiple inputs can can be linked further. This makes it possible for one Measure to use multiple inputs at the same time, even if they are not particularly meaningful individually. Leading on average at a faster response time. Furthermore can use more complex logics for situation interpretation, that are only possible by linking a lot of data (Traffic volume, speed and local density at the cross section and on the measuring cross section on or behind, Trend factors, possibly environmental data), better with fuzzy logic to get managed. Because of the soft mindset of fuzzy logic it is possible instead of a rigid binary decision (Jam or no jam on a cross section) a soft one To determine transition in the form of a probability (e.g. the probability of a jam at this cross-section is 70%). That has the Advantage that this result with a correspondingly predetermined Threshold value can be evaluated so that a reliable display recommendation can be made.

In addition to the vehicle speed, which is determined at both measuring points and in general as a smoothed mean (v) for the respective measuring point is processed, the traffic volume (Q), which also traffic volume is called, and the traffic density (K) is used. The Traffic strength gives the number of vehicles on a measurement cross section on, based on a unit of time, for example one hour. The traffic density is a measure of the number of the vehicles related to a certain section of the route. It is operated with a so-called local traffic density, the the number of vehicles on the measurement cross section relates and takes into account the corresponding speed. The traffic density is the quotient of the traffic volume and the average speed (K = Q / v).

The traffic parameter speed density difference vk-D average speed is calculated from the local traffic data and the traffic density of two neighboring cross sections (Measuring points) according to that specified in claim 1 Formula. The first term of the speed density difference refers to the measuring cross section MQi, the second to the downstream measuring cross section MQ (i + 1). To the traffic sizes different measurement cross sections compare to , they are each set to the adjustable maximum values related to the traffic quantities of the cross sections (max. free speed and max. Traffic density). Is the traffic condition undisturbed at the measuring cross-section, i.e. the speed is not small and the traffic density is not great then the corresponding term is very small in the range Values. If there is an unstable traffic condition on the measurement cross-section, i.e. the speed is low and the traffic density large, the value of the term concerned increases on. Conclusions can thus be drawn from the difference between the two terms be pulled to the current traffic condition.

The trend factor (FT) is used as an indicator of a fault. It monitors the inflow and outflow of vehicles in the measurement section (MA) that have a specific Route length, for example 3 km, and is formed by the two measuring points (MQi and (MQi + 1)). In the event of a critical traffic situation, more vehicles drive into and out of the measuring section, the trend factor (FT) thereby increases exponentially. The calculation of the Trend factor is based on the generally unsmoothed traffic volumes, i.e. the heavy traffic on the two cross sections. This will result in higher accuracy and faster Response achieved. To the influence of measurement errors too decrease, the trend factor is only based on the last measuring interval, that means a period of for example 30 minutes.

The third traffic parameter, the traffic strength trend (QTi), serves to assess the dynamic situation development. The calculation is based on the generally unsmoothed recorded traffic data. The traffic strength trend will also considered at the two measuring cross sections.

These three traffic parameters are the input data for the Fuzzy logic. This brings the input variables, that of two neighboring measuring cross-sections originate from one by rules defined knowledge base in connection and leads the Probability of a critical traffic situation, for example a major accident.

The input variables of the fuzzy logic are influenced by many factors dependent, in particular on the measuring point distance, the route geometry, i.e. Incline or descent, environmental conditions, such as Wetness, snow, black ice, day or night, and possible others Influences. So the influences are not just stationary, but also of a dynamic nature. Therefore, in continuing education the traffic parameters of the method according to the invention calibrated so that the fuzzy system inputs (Traffic parameters) always independent of external influences can rate right away. The sizes are dependent on this their past values dynamically calibrated.

To reduce the effort for calibration to detect a critical Minimize traffic situation will be automatic the trend factor and the speed difference calibrated. For this purpose, this traffic data becomes a calibration factor for the speed density difference and a calibration factor formed for the trend factor and in a calibration facility, between the traffic data preparation and the fuzzy editing is arranged with the current one Relating traffic parameters. The actual Velocity density difference is determined by the velocity density difference calibration factor and the respective one current trend factor divided by the trend calibration factor.

As already mentioned, the speed density difference depends on environmental conditions such as wet, fog, day / night etc. This fuzzy input variable is therefore evaluated by the dynamic calibration factor. The value of this factor can be used as a threshold for the speed density difference above which there is a high probability of a critical traffic situation (disruption). The calibration factor is only calculated if the speed density difference is below a certain threshold, for example 0.3. The factor is made up of the mean, the standard deviation from the speed density difference and its defined threshold. The calculation of the mean value and the standard deviation is only carried out on the basis of the relative maxima of the speed density-difference curve. vk_D medium = αvk_D + (1-α) vk_D middle age σvk_D = α · (vk_D-vk_D medium ) 2nd + (1-α) · σvk_D old vk_D middle age = vk_D medium σvk_D old = σvk_D Calibration factor vk_D = vk_D medium +2. σ vk_D +0.35 PSG PSG = normalized expression (positive very large)
with α = 0.05 (adjustable)

The current speed density difference is shown by this Divided calibration factor. To calibrate the Trend factor, the characteristic value of the trend factor is sought, which you consider "small". This characteristic value becomes like this defines that it includes the set of all values of the trend factor, their relative cumulative frequency below one Threshold. A frequency table is introduced whose classes are defined according to the table will. A class is a defined range of values of the Trend factor, with all classes together covering the entire value range describe the trend factor. For every measuring interval the current trend factor is assigned to a class that the respective class is then incremented. For every interval can be used to determine the measured value for which the relative Total frequency below the specified threshold lies.

The classification is more graded for very small and large values, a finer gradation is chosen for the important calibration range: class 0 1 2nd 3rd ......... 36 37 Characteristic value range -1..0.15 > 0.15 > 0.225 > 0.275 ......... > 1,975 > 2,025 F _T ≤0.225 ≤0.275 ≤0.325 ≤2.025 ≤2,075 Classification

The calibration factor is then calculated: KFT = characteristic normalized expression "small" of the trend factor

The current trend factor is determined by this calibration factor divided.

The method according to the invention for detecting critical traffic situations is used in a special embodiment of the invention for accident detection. From the traffic parameters: trend factor and traffic strength trend of the first measurement cross-section, a crowd formation is recognized in a crowd detection and a pulse probability quantity is derived. In a preliminary accident investigation, an accident criterion is derived from the traffic parameter traffic strength trend of the second measurement cross section and the pulse probability variable with the help of the fuzzy decision, which together with the trend factor and the speed density difference enables an accident detection.
In addition to the decision criteria of speed density difference and trend factor, the traffic parameters, traffic strength trend at the measuring point MQ1 and at the measuring point MQ2 are used for the fuzzy accident decision, with which a preliminary investigation for an accident is carried out.

A bulk detection is carried out. A crowd hires Vehicle collective with high traffic volume and traffic density represents which enters the measuring section.

The traffic parameter used for accident detection Trend factor leaves two interpretations for very large values to. There is an accident, i.e. about one longer period there are more vehicles in the measurement section retracted as extended, or is in the measuring section Bunch retracted. A bunch is like a density wave, such as in the event of a sudden cancellation of a Bottlenecks arise. To differentiate between these two cases as stated above, a bulk detection is carried out. As an input variable of the fuzzy logic, the traffic volume trend, the pulse probability in the previous one Measurement interval, and the trend factor used. As an output variable is a value for the probability of one Pulks are available in the considered measurement section.

With the preliminary accident investigation, the quantities traffic volume trend, old accident probability and pulse probability concluded on the possibility of an accident. The possibility of a malfunction is determined by the output variable Accident criterion represented. Is this value high, so the preliminary investigation indicates an accident.

In addition, the traffic volume trend on the downstream is now also decreasing Measurement cross section from at a high value for an accident criterion, so is the probability of an accident very high. With increasing traffic volume trend on the downstream measuring cross-section decreases the possibility of an accident, as well as when the pulse probability increases. A The exception is the case if a Disorder was very likely. Here is the accident criterion essentially independent of the pulse probability and the traffic volume trend, because in the case of the already In the last measurement interval, a fault identified can both the pulse rate as well as the traffic volume trend rise again. Accident detection is the decision level which ultimately results in the probability of a Accident leads. Depending on this size, a Warning activated on the display cross-section.

Using a fuzzy rule base, the variables become: accident criterion, Trend factor and speed density difference that The probability of an accident is derived, as already explained. With a very large positive speed density difference there is most likely an accident. The bigger (positive) the difference in speed density, the more likely is an accident. With a growing trend factor increases with a positive speed density difference Probability of an accident even more. At a large accident criterion, the trend factor has more influence. With a smaller accident criterion, i.e. interpret the characteristics the speed density difference is not decisive for an accident alone because in this case they are safer than the trend factor. The result evaluation is based on on the likelihood of an accident, a recommendation to display, e.g. Traffic jam warning, for the variable message signs derived and causes the display.

Fig. 1 eine prinzipielle Darstellung für das erfindungsgemäße Verfahren,

Fig. 2 für eine Kalibrierung und

Fig. 3 für eine Störfallerkennung.

The method according to the invention is briefly explained again with the aid of the drawing. Show

1 shows a basic illustration for the method according to the invention,

Fig. 2 for a calibration and

Fig. 3 for an accident detection.

The AS autostraße, here as a freeway with e.g. two tracks in shown a direction of travel, has two measurement cross-sections MQi and MQ (i + 1), which are arranged at a certain distance are and form a measuring section MA. With the traffic sensors VS, e.g. Vehicle detectors, for example can be formed by induction double loops Traffic data VD recorded and a traffic data preparation VDA supplied. Speed is used as traffic data v, the traffic density K and the traffic volume Q detected and processed further. In the traffic data processing VDA the traffic parameters: speed density difference vk-D, the trend factor FT and the high-traffic QTi and QTi + 1 determined separately on the measuring cross sections MQi and MQi + t and fed to fuzzy logic for further processing. The fuzzy processing device is called FUB. The probability quantity formed there, as already explained above Flat share for a critical traffic situation is evaluated in the result evaluation facility EBE on the basis of a Predeterminable threshold value SW evaluated to a control signal SG, for example as a display recommendation for a variable message sign To generate VWZ.

2 is the calibration already described above shown schematically. In a facility for calibration factor formation KFB becomes the traffic data VD or traffic parameters vk-D and FT to form a calibration factor for the speed density difference KFv and one Calibration factor used for the trend factor KFT. These factors are fed to the calibration device KE, in which the traffic parameters speed density difference and trend factor thus calibrated and / or as calibrated Parameters vK-D; FT the fuzzy editing FUB for the further processing already explained can be supplied.

In Fig. 3 the accident detection is shown schematically. In the bulk detection PE, the input variables become a trend factor FT and traffic volume trend QTi at the measurement cross section MQi one Pulse probability quantity PWG using fuzzy logic derived. This pulse probability quantity PWG is in a preliminary accident investigation STV with the traffic parameter Traffic strength trend QT (i + 1) of the measurement cross section MQ (i + 1) considered and derived an accident criterion STK. This criterion STK is used together with the trend factor FT and the velocity density difference vk-D is considered to be to be able to close an accident. This is with the accident detection STE indicated. As explained above, the Accident detection STE to an accident probability quantity SWG closed in a subsequent earnings assessment facility EBE is treated further.

Claims (5)

1. Method of sensing traffic and detecting traffic situations on roads (AS), preferably motorways, having measuring points, designated measuring cross-sections (MQ1, MQ2...), set up for the purpose for vehicle detection using traffic sensors (VS) and a traffic data processing arrangement (VDVE) for traffic control, determining at regular intervals at the measuring points (MQ1, MQ2, ...) traffic data (VD), such as vehicle speed (v), traffic intensity (Q), i.e. the number of vehicles at a measuring cross-section based on a limit of time, and traffic density (K), i.e. the number of vehicles based on a specific section of road, and forming in a traffic data processing system (VDA) traffic parameters determined therefrom, furthermore, two adjacent measuring points, for example measuring cross-sections MQ1 and MQ2, or, as expressed generally, measuring cross-sections Mqi and MQ(i+1), forming a measuring section (MA) of a specific length (l),
   characterized by the following features:
   the following traffic parameters are formed from the traffic data (VD) of two such measuring points:

   a) the speed density difference (vk-D) according to the following relationship: vk - D = v fi-v i v fi + k i k maxi - v fi+1-v i+1 v fi+1 + k i+1 k maxi+1 where

   Vfi, VF(i+1) :

   Adjustable maximum value of the speed at the measuring cross-section Mqi, MQ(i+1)

   kmaxi, kmax(i+1) :

   Adjustable maximum value of the traffic density at the measuring cross-section MQi, MQ(i+1)

   ki :

   Traffic density after the measuring cross-section MQi

   k(i+1) :

   Traffic density before the measuring cross-section MQ(i+1)

   vi, v(i+1) :

   Average speed at the cross-section MQi, MQ(i+1)

   b) a trend factor (FT), which is formed continually from the ratio between the traffic intensities (Qi/Q(i+1)) of the first and second measuring cross-sections (MQi, MQ(i+1)), but determined during a given period (t) in the minute range,

   c) the traffic intensity trend (QTi, QT(i+1)) of the respective measuring cross-sections (MQi, MQ(i+1)), the trend being derived on the basis of the function of the traffic intensity (Q) over the time (curve Q(t)) from the increase of the tangent to the curve,

   these three traffic parameters (vk-D; FT; QTi and QT(i+1)) are processed in a fuzzy logic (FUB) for the detection of a critical traffic situation in the measuring section (MA) being considered and are fed as probability variables (WG) to a downstream result assessment arrangement (EBE), in which control signals (SG) for variable message signs (WVZ) are formed in dependence on adjustable threshold values (SW).
2. Method according to Claim 1, characterized in that the traffic parameters of speed density difference (vk-D) and trend factor (FT) are dynamically calibrated in dependence on their past values, a calibration factor (KFv) for the speed density difference (vk-D) and a calibration factor (KFT) for the trend factor (FT) being formed (KFB) from the traffic data (VD) and in that in a calibration arrangement (KE), arranged between the traffic data processing system (VDA) and the fuzzy processing (FUB), the current speed density difference (vk-D) is divided by the speed density difference calibration factor (KFv) and the respectively current trend factor (FT) is divided by the trend factor calibration factor (KFT).
3. Method according to Claim 2, characterized in that, for calibration of the speed density difference (vk-D), the latter is assessed, the value of the calibration factor (KFv) for the speed density difference being a threshold value for the speed density difference (vk-D) from which there is a high probability of a critical traffic situation.
4. Method according to Claim 2, characterized in that, for calibration of the trend factor (FT), a characteristic value of the trend factor estimated to be "small" is defined such that it comprises the set of all the values of the trend factor whose relative cumulative frequency lies below a threshold value, a frequency table being formed with a plurality of classes with defined ranges of values of the trend factor and the current trend factor being assigned to a class in order to determine the calibration factor (KFT) from it.
5. Method according to one of Claims 1 to 4, characterized in that a disruption is determined and displayed as a critical traffic situation, the trend factor (FT) and the traffic intensity trend (QTi) of the first measuring cross-section (MQi) being used to detect bunching (PE) and to form a bunching probability variable (PWG), the relationship of which with the traffic intensity trend (QT(i-1)) of the second measuring cross-section (MQ(i+1)) is established in order to derive (STV) a disruption criterion (STK), the trend factor (FT) and the speed density difference (vk-D) and also the disruption criterion (STK) also being used to detect (STE) a disruption and to form a disruption probability variable (SWG).

Images