BE1019524A3 - SYSTEM AND METHOD FOR TRAVEL TIME MEASUREMENT. - Google Patents

Abstract

The invention relates to a system for accurately measuring, aggregating, predicting and publishing traffic information, characterized in that the road network of a digital map is segmented into road segments of a fixed length on which probes of at least 2 GPS positions per vehicle are matched. The invention further relates to a method for this, comprising the following steps: a) re-segmenting a digital map into road segments of a fixed length; b) matching GPS positions of probes on the road segments; c) deriving traffic information by segment.

Description

System and method for travel measurement

The invention relates to a system for accurately measuring, aggregating, predicting and publishing, among other things, travel times, speeds and delays of road traffic based on Floating Car Data. The invention also relates to a method for performing such a travel time measurement and deriving related traffic information.

Floating Car Data (FCD) is a technology that monitors traffic (anonymously) using GPS (Global Positioning System) or, less accurately, via the mobile phone (GSM). A vehicle that is equipped with FCD technology, also called a probe, can be followed accurately in time and space via GPS. With satellite positioning it is possible to collect a multitude of vehicle data. In addition to the origin-destination location, the distance of a journey and the travel time can be determined. By linking the time-space data to a digital road map, the translation can be made to the road-related traffic information.

There are various technical systems on the market that determine the locations of vehicles with the help of the network of mobile phones or GPS, and use this to get a picture of the traffic situation.

US 6,650,948 (Applied Generics) describes a "vehicle traffic monitoring system" operated via mobile telephony, where the geographical positional data is mapped on the discrete road components that are part of a road network.

US 6,879,907 (TrafficSoft, formerly Estimotion) describes a traffic system based on oriented road sections, wherein a spatial representation of a road network is first transformed into a network of spatially interdependent and interrelated oriented road sections, with which the traffic data collected is then associated.

The applicant initially also worked on the basis of moving GSMs, but this has now been discontinued. On the one hand it turned out that the accuracy was inadequate (GSM cells are inaccurate, coverage decreases with slower traffic, ...) and on the other hand the number of GPS connected vehicles had increased considerably.

Existing systems typically map the collected data on (part of) the road network of a digital map. The concepts of route planners and digital road maps are used, being the segments between two nodes as defined and customary in geographic information systems. A road segment is a vector between two points where the road segment changes its attribute or where an intersection is created with other road segments, such as a side street, an intersection, an exit, etc.

A disadvantage of mapping on unequal and longer segments is that the individual observations are spread out over the entire long segments.

Other existing systems use roadside measurements (eg countuses and cameras) that have been installed in the roads to count the number of vehicles that pass by somewhere. These measure local speeds (point speeds). Based on this, a prediction is made of the travel time on a route. A disadvantage here is that this is a prediction. Through the present invention, the traveling time is available just the other way around and, based on that, the number of vehicles can be derived if necessary.

It is an object of the invention to provide an answer to the question why people are now standing still in a traffic jam. The travel time information of the measurement system is accurate to within 50 meters.

The measurement system of the present invention is based on principles of "Traffic Flow Theory" but still includes some important differences.

In traffic flow theory, connections are made in a mathematical way between the density of vehicles on a road section, the speed of the vehicles and the traffic volume. Crucial in this theory is that there is a relationship between all these parameters and that further work is done based on the known traffic volume. Because the traffic volume is always required, the development and application of theory in practice always involves traffic volumes that work with roadside systems. Roadside systems measure the number of vehicles and the speed at a point at a certain location on the side of the road.

The measuring system of the invention is based on following a large sample of the moving vehicles. As a result, a statistical sample of the total traffic population is measured and not the exact number of vehicles. The invention has further expanded the traffic flow theory to take into account that the work is not based on total traffic volumes, but on a statistical part of the traffic flow. Furthermore, no information from fixed locations along the road is used, but information from moving vehicles is used.

Traffic jams are big business. The current invention measures travel times, and makes that information available to navigation brands that process the information live and forward it to their navigation boxes via mobile internet. To receive the info, a recent navigation system with a "live" subscription is required. In exchange, the user receives very accurate arrival times, or in the best case a faster route to his destination, even if it is abroad. An electronic agenda can even indicate precisely up to 10 minutes exactly when one has to go to an appointment to have less traffic jams.

Via the measuring system of the invention, it is possible to know at any time how quickly people are driving on the highways and the underlying roads. Information is available every 50 meters, and the situation can be monitored literally minute by minute, because so often the monitor vehicles transmit their GPS position via mobile internet. The system is based on Floating Car Data technology. Mathematical algorithms refine existing digital maps (for example from TeleAtlas and Navteq) so that the information can be projected every 50 meters and GPS positions can be converted into travel times. More advanced navigation systems can handle this accuracy.

Small segments mean that traffic can be assumed to be stationary within a segment. This means that vehicles maintain a constant speed within a segment, so that there is no longitudinal averaging or simplification by working with longer segments.

The object of the invention is achieved by the system and method of the appended claims.

Existing floating car measuring systems do not work with short segments of fixed length. Furthermore, in other measurement systems the published value on a segment is based on the individual observations on this segment. This does not allow to take into account the intrinsic longitudinal and temporal dynamics of the traffic flow. In the present invention, when publishing the traffic state on a segment, not only recent recent observations on this segment are taken into account, but also all observations on upstream and downstream segments. The link between the segments is therefore taken into account to take into account the build-up of traffic waves and the evolution of traffic flow through the traffic network.

Further advantages of the system of the invention are:

Some systems use the travel times that are measured by probes at a certain point. This is not an average travel time, but only contains local speed information. The current system measures travel times between different GPS points and no speed of the vehicles themselves; this requires a route calculation between the monitored GPS positions and we also gain insight into detail between the roads in exchangers.

Probes that are not representative typically affect the published traffic image. The current system includes tailor-made filters for filtering away, among others, stationary vehicles on a traffic lane, identifying stop zones (eg gas stations), black vehicle list with faulty, suspect probes.

Digital cards with different lengths lead to different averages. The current system does not have this problem because the segments have a fixed length as much as possible.

Fittings in route network

Scalable calculation by lookup instead of calculations Systems fail in the origin of data

Difference in road users (eg taxi, trucks, fast drivers, ...) versus difference in traffic regime (free traffic, congestion, ...): We take this more by attaching representativeness to the probes. Blocking of a road and infinite travel time

Longitudinal impact on the traffic image (by relating successive segments upstream and downstream)

Properties of jam growth

Include number of probes (representativeness) in file determination Historical relationships between different sections (congestion mechanism) HB patterns and track analysis Accuracy to less than 1 m.

The historical data from the probes are combined with current or journalistic data from deviating situations (incidents, bad weather, etc.), possibly in combination with systems that measure information along the road (countuses, telradars, cameras, Bluetooth vehicle registration) , license plate recognition, ...).

The information collected can be used for traffic information and dynamic traffic management, in traffic models and for policy purposes.

The source information consists of exact GPS positions required for the measuring system to work. The most important source for supplying GPS positions via mobile internet are the professional vehicles such as taxis, courier services, trucks and so on. They often work with a dispatch system with navigation, and mobile internet on board. In addition, every user of Coyote radar detectors (around 45,000 in Belgium) also passes on his GPS position. This is done completely anonymously and the data cannot be reduced to one driver who is, so to speak, followed day and night.

This automatic information must suffice to see the traffic jams and to measure them accurately. However, as a user you also want to know the cause of a traffic jam, and that "journalistic data" comes from texting notifications at radio stations or texting from people calling a traffic line. Finally, the government also provides some information about road works.

The system also takes into account, when calculating travel times, historical data and file info that has been kept since the start of information gathering.

The technology of the present invention works on a digital map where road segments are divided up to a length of up to 50 meters. This is an essential part of this technology that allows the dynamics of congestion waves to be taken along. Moreover, this avoids the smearing of non-stationary traffic on longer sections.

GPS positions of moving vehicles are used as input for this system. An important part comes from professional vehicles that are equipped with "track and trace" systems and that transmit GPS data in real time. For example, 15 different “track and trace” suppliers send data on average every minute. This fleet contains around 80,000 vehicles that cover a relatively large number of kilometers. This concerns both heavy and light freight, vans, lease vehicles, taxis, ...

Another part of the fleet comes from feedback from connected consumer applications. For example, moving Coyote systems transmit their position every minute. In this way smartphone applications also link locations of moving vehicles.

For one vehicle, the range of GPS positions on the segmented digital map is matched. From this, the travel time of the vehicle between the GPS positions is assigned to the different road segments as an individual observation.

In the real-time system, an average road image is built on this basis and traffic information is generated. These individual observations per segment are further processed to calculate the requested travel times and speeds (average, median, variation and different percentiles) per period (eg between 8 and 9 hours) over the different days.

The fine resolution of the measuring system of the invention offers considerable advantages because it takes the extra detail of file construction with it. Moreover, dynamic start and stop waves are not averaged over longer road segments.

A preferred embodiment of a method according to the invention for accurately measuring, aggregating, predicting and publishing traffic information, as represented in Figure 1, consists of 6 blocks: 1. Digital map preprocessing

In the first process of this system, a digital map (eg from Navteq, Tele-Atlas, ...) is pre-processed. The following steps are discussed: - Selecting the network that is used to collect traffic. This means that lower order roads, for example, are not selected.

- Roads are converted into targeted line sections where the direction of the line segments corresponds to the direction of travel. Roads with double direction of travel are deduplicated into two separate vectors.

Nodes are defined at the start and end of each vector. Different vectors are concatenated (brought together in 1 vector) if the traffic has no other choice to drive into the selected network than the next vector. In this way the network comprises a minimum number of one-way vectors in which the traffic can change from one-way only. These concatenated vectors are separated from each other by "super" nodes.

Segmentation of a digital map (eg Navteq, Tele-Atlas or open source map) based on re-segmentation of vectors between two "super" nodes. The segmented road segments have a constant length, preferably maximum 50 meters, or between 40 and 60 m, between 30 and 50 m, between 20 and 30 m, in any case maximum 100 m. The segmentation allows, among other things, to be able to quickly match a GPS position of a vehicle to a segment.

There is a feedback loop that indicates errors in the digital map based on, for example, too low coverage, many aborted trips, ...

2. Collect and process GPS data

Raw GPS logs from "track-and-trace" systems are then processed in real-time. Other possible linked data include a time stamp, vehicle type (truck or not), GPS signal accuracy, ... The system preferably receives two GPS positions every 1 to 3 minutes from one vehicle. Delivering GPS positions faster than every minute means that many GPS positions must be processed and that the calculation time increases considerably; longer than every 3 minutes the system produces a slowness: it takes at least 3 minutes between the observation and the publication of the traffic information.

Belgium is, for example, subdivided into 3000 km of motorways, partly with measuring loops, which the Government uses to count vehicles and to project information onto the signs. With the current measuring system, 60,000 km of roads can be used. Also exits and smaller roads, sometimes in the city. Preferably, roads are chosen where the traffic function is the most important. The system can have at least 2 percent measuring vehicles on the roads at any time. That 2% provides GPS data for the travel time calculation.

3. Map matching and travel time samples

In a first step, GPS positions are matched on the segmented digital map. This leads to candidate sections of road where this vehicle drives. For example, with a GPS coordinate it is not necessarily clear in which direction the vehicle is moving or if this vehicle is above or below a bridge. This also happens for the second GPS position.

After that a route calculation takes place between the possible candidate segments. This ensures that possible routes are calculated between the different candidate segments. The GPS positions and the routes are selected based on the distances of the GPS to the candidate segments and the optimal travel time on the calculated routes between the two GPS locations (in relation to the time difference between the two time stamps).

4. Other qualitative and quantitative data

Adding point measurements (countuses, counting cameras, radars, ...), trajectory measurements (eg based on blue tooth, automatic number plate recognition, ...), adding community feedback and / or operator input, journalistic status of traffic (eg accidents, road works, incidents, ...).

5. Traffic condition estimation

Based on pre-processed information, the traffic image is calculated via traffic engineering and statistical algorithms as known from the "traffic flow theory".

This means that estimating the average traffic image on a segment not only takes into account the observations of individual vehicles on this segment (preferably in the last half hour), but also the traffic image or the observations on sections upstream and downstream. For this, the system bases itself on the traffic flow theory that determines how the traffic image in free traffic (no congestion) and traffic congestion (but congestion) moves over a road over time. These congestion waves congest the direction of travel (so the downstream sections are relevant in the past) and follow the direction of travel for variations in traffic volume and speed changes during smooth traffic (in which case upstream sections are relevant).

This estimate makes explicit use of congestion waves and this is therefore only possible with short segments of less than preferably 50 meters. This estimate takes into account the value of the speed in up and down sections, but also (the variation of) the number of observations.

6. Publish

The traffic image can be made known to the user in various ways. For example, the following information is published:

Travel times (Route times, delays, ..._) instantaneous and predictive

Traffic jam locations

Traffic jam and traffic reports

Based on the evolution of our coverage (number of observations), an estimate of traffic volume is also made.

File length, file pressure and file cost

Heterogeneous distribution of traffic flow and user dependencies (so different delays for trucks / passenger cars)

Speed distributions as offline outcome: we can estimate the speed distribution on a segment and thus see where, for example, driving is too fast.

Publishing is done almost immediately (online) or offline (for later analyzes, for example).

The measuring system does not use a database as such. This would only delay the processing. The map definitions are not stored in a database, but are exported to CSV format and read in by the system.

The measurement system of the invention is completely transparent via an internal web font end. Hereby the following information is available:

For each GPS provider: percentage of the incoming samples used; which GPS positions are blocked by the filters and why (the GPS positions can be seen as KML in Google Earth); what the current trips are: raw data, calculated route, derived speed.

Total aggregated speed on the base map as KML.

No black box: in case of unexpected results, a drill-down approach leads to the GPS samples that cause the published speeds.

Floating Car Data

The measurement system of the invention follows professional vehicles connected via the mobile internet that directly transmit their GPS positions.

Other systems, such as TomTom, also follow all mobile phones that are "on the move" at that time and are calling. They watch how fast they move from one cell phone to another. The disadvantage is that such a tracing via mobile phone only works if you actually call. The problem is that a position that you obtain in this way are really large zones. With GPS positions, there is an accuracy of up to 10 meters. The slower the traffic, the more positions per km, because the vehicles send information every minute. At times when it really matters (file), the information becomes more relevant. For mobile phone positions, the opposite is true: before you can drive from one mast position to another in a traffic jam, a lot of time is left (about 6 minutes). In addition, the mobile phone data of passing trains must also be properly filtered out.

Typically a position is passed every minute. This amounts to information every 2 km (for a 120 km / h motorway).

Filtering probes

Correct filtering of the probes is an important step in quality. Via the web frontend, all probes from the different providers can be tracked individually and linked back to their input in the measurement. Real-time tracking of all probes is possible for every GPS provider.

One of the filters is a minimum mapped speed of 2 km / h, for example to filter out maintenance vehicles.

Therefore, the probes are assigned representativeness properties for filtering out non-relevant probes.

Quality method

The measuring system includes an internal working method that detects and corrects systematic errors within the collection technology. In the past, this has already led to various internal digital map improvements, identification of black list vehicles (eg excluding site vehicles from the fleet), etc ... This internal follow-up is monitored from a track list.

In addition, there is also a specific quality method from the global end product that, in addition to the collected travel time data, also includes journalistic traffic data. This process is based on the QKZ method developed within car manufacturer BMW and where published information is compared with both user experiences (eg test drives) and other available measurement data. Typical output indicators that are discussed here are false positives (which fake messages / info was distributed), false negatives (which info was not distributed), delays in the process, ...

Both proactive internal processes are supplemented with reactive follow-up of user feedback in order to continuously maintain the quality of the internal measurement system.

Accuracy

The system works internally with a distance of millimeters in a time of nanosecond. The "base map" has an accuracy of 50m. Accurate extrapolation is made within the 50m segments. The theoretical precision is at less than lm.

Search for traffic bottlenecks ("measuring is knowing")

With the measuring system of the present invention, structural bottlenecks on the roads, among other things, can be detected, thus enabling a more efficient approach to such problems. For example, a measurement carried out on a certain route shows that 60 percent of the traffic turns towards B and 40 percent continues towards K. But the lanes are unfortunately arranged differently. There are three lanes in the direction of K and two in the direction of B. It is therefore certain that it is stuck every morning. By simply adjusting the number of lanes, you can already create a major impact.

Prediction of traffic jams This can be done in three forms: 1. Prediction in the long term (week or longer): This only requires historical data: which route do people most often take when they take the A-B route and are on the road at 7.30 am on a Friday. Wednesday and Friday are the quietest mornings, for example. Holiday periods are also included.

2. Medium-term forecasting: for 3 to 4 days, the system also takes weather forecasts into account. On rainy days, traffic jams are on average 30 percent longer.

3. A short-term forecast: between now and 2 hours. Then the system uses all data from 1 and 2, plus all files that are now there. There was an accident half an hour ago, and that will only be resolved within an hour.

Measuring traffic waves

Traffic jams move against the driving directions. Slow traffic thus has a predictive character. The coverage (number of observations) also moves in waves, albeit with the speed of traffic: 10 vehicles that are now running at 60 km per hour will be all ten one kilometer further within the minute. A combination of these two information means that information from up and down segments can be taken into account when determining the average traffic image (speed, travel time) on a segment. The traffic situation on a segment is therefore determined not only on the basis of the available observations in the recent past on this segment, but also on these upstream and downstream segments, although a time shift is taken into account there (e.g. with a golf at 20 kilometers per hour the system must look 3 minutes in the past when considering a segment 1 kilometer further (20 km / hour = 1 km / 3 minutes).

The invention will now be explained in more detail with reference to the following detailed description and figures.

Figure 2 shows an individual route of a single vehicle The GPS data is completely anonymous, but one can pick out one vehicle and see which route it follows, and how quickly it advances on that route. We get a GPS position every two minutes, so the measurement system knows how fast the vehicle was driving on average on that road between two points.

Figure 3 shows a TX diagram of the E313 Herentals Industrie route to Antwerp East.

A TX diagram is the standard way in which a situation on the road during a period (here the morning rush hour) is depicted. The horizontal axis is the time. The vertical axis is the distance traveled. The vertical axis on the right indicates the speed: the darker, the more traffic (from green to red). A TX diagram can almost indicate when it is best to start the trajectory. The white lines are cars that leave at the bottom and, depending on the time, whether or not they get stuck in traffic. The more vertical they go up, the better they can drive on. If the lines run horizontally, they lose time in the traffic jam. The diagram shows very clearly how the traffic jam hits back in different waves against the direction of travel (traffic jam goes backwards, from top to bottom, while the white lines, the cars, drive from bottom to top). If you leave at 6 am, you drive up vertically, if you leave fifteen minutes later, you can end up somewhere in the middle of a traffic jam, in a tail of the traffic jam (a yellow or red piece). One wonders then: how can there be traffic jams here, nothing has happened here. That is the harmonica effect. The car will also come to the rear of the line when it is affixed. People drive away again, but in this way the traffic jam moves backwards, against the direction of travel. The traffic jam always moves (backwards) at a more or less constant speed: 17 km / hour. Although someone left only 10 minutes later than their predecessor (the line at the bottom is close to the other line), he has lost much more time along the way (lines are further apart as you go up). On the basis of these real-time graphs, the electronic agenda of the future could say: "No, you have an appointment there at that time, then you must leave now, because an accident has happened here and there." A navigation system could even get a say in the location where you organize the meeting.

Figure 4 shows a TX diagram when performing work on the motorway This TX diagram shows the placement of noise barriers: a traffic jam from the E313, all the way around the Antwerp ring road, to Haasdonk on the El 7 Suppose that people leave in Haasdonk at 1 pm. They go up fairly quickly, steeply. Just before Antwerp West one gets to the orange part, and one has a fairly strong traffic jam, one starts to drive more horizontally. Just before the really dark part, you hit Antwerp South. But if, for example, you leave at 2 p.m., you will already be in a traffic jam for Antwerp West, traffic will slow down in Zwijndrecht, and you will go very slowly (horizontally) through the Kennedy tunnel. Then you also end up in the darkest part. If you were to leave in Haasdonk at 4 pm, you would be too late, because you would then drive straight into the traffic jam and you would have to go straight through the traffic jam (which goes backwards). It is noticeable that the delay that occurs on the E313 has a delaying effect on the Antwerp Ring towards the Netherlands. This effect is fairly limited in the morning (the traffic then goes mainly to Brussels and to Ghent), but from the afternoon onwards the traffic volume on the Antwerp ring road increases and the work activities have a greater effect.

Figure 5a shows a TX diagram of file formation with a coverage (observations) as represented in Figure 5b

The horizontal changes in coverage (for example between km 16 and 18) have to do with on and afflicting traffic: at km 16 a relatively large amount of traffic runs towards the Brussels ring road. Extra traffic will be added at km 18. The exit at km 20 is noticeable in a similar way.

It is also striking how the variation in coverage evolves with the speed of traffic, for example between 8 am and 8.30 am between km 6 and 14; considerably faster traffic after 10 hours between km 0 and km 4.

It is also striking how the coverage falls relatively in traffic. One vehicle at 120 km per hour contributes to coverage in 40 segments in 1 minute. A vehicle that only runs at 6 km per hour, however, only passes 2 segments in 1 minute.

Figure 6 is a representation of traffic jams versus travel time for a specific route.

Figure 7 is a representation of click-on-demand, in which a traffic situation can be displayed at a specific time in a specific zone.

Figure 8 is a representation of the measurement system of the invention via live segments on Google Maps. The color codes indicate the occupancy of the "50m segments". Dark colored segments are busier than lighter colored ones.

Figure 9 illustrates a screenshot of the measurement system of the invention. Dark colored segments are busier than lighter colored ones.

Figure 10 is a representation of the coverage per segment (number of probes) on an aggregated base map as a function of time.

Figure 11 shows the length of the file in road class class.

Claims (18)

A system for accurately measuring, aggregating, predicting and publishing traffic information, characterized in that the road network of a digital map is segmented into road segments with a fixed length on which probes of at least 2 gps positions per vehicle are matched. System according to claim 1, characterized in that the traffic information is allocated per segment. System according to claim 1 or 2, characterized in that each segment is no larger than 100 m, preferably 50 m. A system according to the preceding claims, characterized in that the segments are located between two super nodes that involve a change of direction. System according to the preceding claims, characterized in that the traffic information comprises historical and journalistic data. A system according to the preceding claims, characterized in that the probes are based on Floating Car Data. System according to the preceding claims, characterized in that the traffic image on a segment takes into account observations of upstream and downstream segments. A system according to the preceding claims, characterized in that the traffic image on a segment takes into account the progress of the congestion waves and the progress of (the variation of) the number of vehicles as a predictive factor through time and space. A method for accurately measuring, aggregating, predicting traffic information, characterized in that the method comprises the following steps: a) re-segmenting a digital map into fixed-length road segments; b) matching GPS positions of probes on the road segments; c) deriving traffic information per segment. The method of claim 9, wherein the traffic information is published. A method according to claim 9 or 10, wherein the fixed length of a road segment is no greater than 100 m, preferably no greater than 50 m. The method of any one of claims 9 to 11, wherein at least 2 consecutive GPS positions of the same vehicle are used A method according to any one of claims 9 to 12, wherein in step b) and / or c) also other data, such as journalistic data, for example, is matched on the road segments and / or contributes to the derived traffic information. A method according to any of claims 10 to 13, wherein the published traffic information may comprise the following info: - instantaneous and / or predictive travel times such as, inter alia, journey times, delays. traffic jams and traffic jams such as, for example, file location, file length, traffic jams, traffic jams, traffic volume, heterogeneous distribution of traffic flow, and user dependency, such as delay for trucks and / or passenger cars, speed distribution The method of any one of claims 10 to 14, wherein the published traffic information is delivered either quasi directly (online) or offline. The method of any one of claims 9 to 15, wherein non-representative probes are filtered out. The method of any one of claims 9 to 16, wherein historical relationships between different segments can be deduced. The method of any one of claims 9 to 17, wherein the traffic image has an accuracy of less than 1 m.