US20120277985A1

US20120277985A1 - Method of analyzing points of interest with probe data

Info

Publication number: US20120277985A1
Application number: US13/504,491
Authority: US
Inventors: James Alan Witmer
Original assignee: TomTom North America Inc
Current assignee: TomTom North America Inc
Priority date: 2009-10-29
Filing date: 2009-12-31
Publication date: 2012-11-01
Also published as: WO2011053340A1; TW201118345A; JP2013509639A; US20120278505A1; TW201115118A; WO2011053337A1; WO2011053338A1; US20120310772A1; TW201126139A; WO2011053336A1; TW201123024A; TW201115172A; EP2494309A1; TW201115498A; US20120271864A1; JP5670464B2; WO2011053388A1; CN102667404A; WO2011053391A1; TW201115119A

Abstract

A method of analyzing points of interest (22) using traces from probe data is provided. The method includes providing a database of a digital vector map (18) configured to store a plurality of traces (1′-14′) representing roads. The method further includes collecting probe data from vehicles traveling along the traces. Then, bundling a group of select traces (2′, 5′, 7′, 9′, 11′) having routes with a common origin (20) and at least one divergence point (24, 1) downstream from the origin (20) and building a database of vehicle maneuvers over the routes. Further, computing average speeds and delay times of a random population of vehicles traversing the vehicle maneuvers. Further yet, computing average speeds and delay times of all vehicles traversing the routes. Then, comparing the computed results from the random population of vehicles with the computed results from all vehicles traversing said routes.

Description

BACKGROUND OF THE INVENTION

1. Field of the Invention
This invention relates generally to methods for analyzing points of interest, and more particularly to methods of analyzing points of interest with Global Positioning System (GPS)-enabled devices.
2. Related Art
Points of interest (POIs) are often analyzed via tabular information, such as via manual research; via directories of restaurants in a chain with their addresses; points supplied by customers, third parties, address lists, and the like, wherein the points of interest are assigned a coordinate (latitude/longitude) via manual research or automated means such as geocoding (the use of a known address and map match to assign an approximate coordinate). Unfortunately, the results can be fraught with errors, such as due to human error; furthermore, not all addresses are available on point lists, may have been changed, or the map itself may have incorrect street names or numbers, leading to incorrect locations. Further, rating of the POIs is typically manual, and thus, generally proves difficult and costly. In addition, the manual data gathered can become dated in a relatively short period of time, thereby rendering the data obsolete and increasingly inaccurate.

SUMMARY OF THE INVENTION

In accordance with one aspect of the invention, a method of analyzing points of interest using traces from probe data is provided. The method includes providing a database of a digital vector map configured to store a plurality of traces representing roads and collecting probe data from vehicles traveling along the traces. Then, bundling a group of select traces having routes with a common origin and at least one divergence point downstream from the origin and building a database of vehicle maneuvers over the routes. Further, computing average speeds and delay times of a random population of vehicles traversing the vehicle maneuvers. Further yet, computing average speeds and delay times of all vehicles traversing the routes. Then, comparing the computed results from the random population of vehicles with the computed results from all vehicles traversing said routes.
Upon comparing the computed results from the random population of vehicles with the computed results of all vehicles traversing the selected routes, statistically probable differences may be discerned. Accordingly, POIs are able to be identified by noting the differences in vehicle behavior over the selected routes.

BRIEF DESCRIPTION OF THE DRAWINGS

These and other aspects, features and advantages of the invention will become more readily appreciated when considered in connection with the following detailed description of presently preferred embodiments and best mode, appended claims and accompanying drawings, in which:

FIG. 1 illustrates a selected bundle of traces to be analyzed from a navigable street network database;

FIGS. 2A-2D illustrate examples of average speeds and delay times of vehicles over a select portion of the bundles of traces of FIG. 1;

FIG. 3 illustrates another selected bundle of traces to be analyzed;

FIGS. 4A-4D illustrate the average speed and delay time of vehicles over two distinct points of the selected bundle of FIG. 3; and

FIGS. 5A-5B illustrate further statistical analysis of data taken at one of the distinct points of FIGS. 4A-4D.

DETAILED DESCRIPTION OF PRESENTLY PREFERRED EMBODIMENTS

In accordance with one aspect of the invention, information is obtained from global behavior of vehicles traveling along a navigable street network, wherein the street network is defined by a plurality of traces. The information is useful to assess specific behavior of the vehicles, and thus, can be used to determine where particular points of interest (POIs) exist along the navigable street network. The POI can be pre-existing, or new. The information gathered can be obtained substantially real-time, and thus, the information is current and reliable. Further, since the navigable street network undergoes dynamic change, the changes that occur can be monitored and processed in an economical manner, without need for manual data gathering. The information can be used to determine the decision patterns of travelers, whether they be utilizing motorized vehicles, bicycles, pedestrian travel, or otherwise. Accordingly, the invention is not limited to assessing the behavior of motorized vehicles.
Referring in more detail to the drawings, FIG. 1 illustrates an example of how a database of traces can be bundled and selected for analysis from a portion of a navigable street network 10, by way of example and without limitation. The bundled traces exemplified have been assigned end points enumerated 1-4. Each of the traces originates from a common location or origin 12 and pass through a common intersection 14. As can be seen some of the traces share travel paths over a portion of their distance, such as exhibited by trace bundles 3 and 4 until they reach a divergence point 16. Accordingly, when gathering data for vehicles traveling along trace bundles 3 and 4, it is important that the geographic extent of the data collected from vehicles traveling along these bundles be obtained far enough along their respective paths to detect the decisions made by the vehicles upon reaching the divergence point 16 in the bundles 3 and 4. In addition, upon reaching the divergence point 16, as can be imagined, vehicles wishing to travel along trace bundle 3 may be delayed at the divergence point 16 for any number of reasons. For example, some vehicles may have to wait for a line of vehicles prior to turning, or they may have to wait for an indicator signal.
In order to assess the travel behavior of the vehicles traveling along the bundled traces 1-4, the maneuvers of the vehicles traveling along the bundled traces 1-4 can be analyzed. As shown in FIGS. 2A-2D, wherein FIG. 2A corresponds to bundled trace 1, FIG. 2B corresponds to bundled trace 2, FIG. 2C corresponds to bundled trace 3, and FIG. 2D corresponds to bundled trace 4, an average speed profile of the vehicles (column L) and an average delay time profile of the vehicles (column R) over the bundled traces 1-4 can be obtained via probe data received from the vehicles. Any suitable statistical program application can be used to generate the averaged information. As can be seen, the information obtained for vehicles traveling along trace bundle 1 indicates that the vehicles turning left come to a stop or near stop and then resume a speed immediately thereafter approximating the speed prior to making the turn. In contrast, the vehicles traveling along trace bundle 2 behave differently than those traveling along trace bundle 1, wherein the vehicles traveling along trace bundle 2 slow slightly from their top end speed while traveling through the intersection 14 and then resume immediately thereafter their top end speed. In further contrast, the vehicles traveling along trace bundle 3 exhibit the slowest average speed at the intersection 14 to make a right hand turn, then speed up slightly, followed by another decline in average speed to make a left hand turn at the divergence point 16. Then, for the vehicles traveling along trace bundle 4, the behavior at the intersection 14 is markedly different from those desiring to travel along trace bundle 3, wherein although the vehicles traveling along trace bundle 4 slow their average speed to make a right hand turn at the intersection, the decline in average speed is not nearly as great as those traveling along trace bundle 3. Accordingly, it is important to be able to distinguish between the vehicles traveling along the traces in bundles 3 and 4 in order to obtain meaningful conclusions regarding their behavior. With the database of maneuvers over the traces 1-4 now constructed, probe data obtained from vehicles traveling the different trace bundles 1-4 can be used to assess POIs attracting the vehicles and likewise, POIs can be populated or verified on the trace bundles 1-4 to bring them to the attention to the users of Global Positioning System (GPS)-enabled personal navigation devices, such as those manufactured by TomTom NV (www.tomtom.com). However, any suitable device with GPS functionality may be used, including handheld devices, Personal Digital Assistants (PDAs), mobile phones, and the like.
In an example of how a database of maneuvers over a selected group of traces on a navigable street network 18 can be utilized, we now refer to FIG. 3, which depicts a database of bundled traces 1′-14′ exiting a common location, wherein the location is represented as an airport 20, by way of example and without limitation. It should be recognized that, for discussion purposes, the illustration is simplified, and that the bundled traces can be as complex as necessary to encompass the area desired for study. Each trace path 1′-14′ is associated with the sequence of maneuvers that it follows upon exiting the airport 20. The sequence of maneuvers can be continued until the trace ends; until the trace returns to the airport 20, such as often occurs with taxis or buses; until the trace extends beyond a predetermined geographical limit, or until the trace exhibits a marked decrease in travel by vehicles, for example. Of course, other constraints can be used to determine the traces and how far to extend analysis thereof.
In our example, we note that starting with the exit of the airport 20, that traces 1′ and 2′ are the only possible decisions for vehicles to travel. Upon study, we learn from probe data received that the vast majority of vehicles leaving the airport 20 continue along trace 2′, and that only slight minority travel along trace 1′. So, for purposes of assessing POIs for vehicles leaving the airport 20, we discount those vehicles electing to travel trace 1′, and continue monitoring probe data from those vehicles traveling along trace 2′. We continue this line of reasoning until there is no one favored trace of travel over another, and by doing so, we learn from probe data that the most favored traces traveled by vehicles are 2′, 5′, 7′, 9′ and 11′, and that upon reaching the intersection (I) at 12′, 13′ and 14′, there is no clear favored trace traveled by vehicles exiting the airport 20. And so, for our specific purpose of vehicle behavior study, we elect to study the selected series of maneuvers (referred to as “route”) of the vehicles traveling the probe traces 2′, 5′, 7′, 9′ , 11′ (referred to as “group”) through the maneuver ending at trace 11′.
In order to determine POIs located along the group 2′, 5′, 7′, 9′ , 11′ of study, and in our example, a POI being represented as a hotel 22, an algorithm is used to compare the behavior in maneuvers (speed through the maneuver, stop time at decision point) between an overall random population of vehicles and vehicles leaving airport 20, referred to as the airport group. If the behavior between the two populations of vehicles diverge such that it is statistically probable that they are different, then a POI can be determined.
As illustrated in FIGS. 4A-4D, by way of example, it can be readily determined that there is a POI (hotel 22) at the maneuver for trace 9′, while there is no POI at the maneuver for trace 5′. The FIGS. 4A-4B show the average speed (column L) and the average time delay (column R) of the bundled trace 5′ for the overall random population of vehicles (FIG. 4A) and the airport group (FIG. 4B). In these figures we add the mean delay time in seconds (μ), and the median or 50^thpercentile delay time, also in seconds (M), for illustrative purposes, In comparing the two vehicle populations, it is apparent that there is little difference in their behavior. Though there is some difference, it is slight, and could be attributed to such things as statistical sampling, differences in vehicle types and behavioral differences due to things that have nothing to do with a POI, such as taxi cabs driving at different speeds or public transportation restricted to certain lanes, for example.
In contrast, the bundled trace 9′, as shown in FIGS. 4C-4D wherein the average speed (column L) and the average time delay (column R) for the overall random population of vehicles (FIG. 4C) and the airport group (FIG. 4D) are shown, it is apparent that there is a substantial difference between the behavior of the separate populations which can be inferred by statistical methods. As can be clearly seen from the dramatic drop in the speed profile of the selected airport group, there is a detected POI approximately 65 meters prior to a divergence point 24 where trace 9′ diverges from trace 8′. This discernible drop in speed, however, does not appear for the overall random population shown in FIG. 4C. Accordingly, these illustrations show the value of comparing vehicles within the specific target group versus the larger random population.
Of course, depending on the nature of the POI, the number of vehicles in the group stopping at the POI could vary. As such, in accordance with the invention, additional statistical analysis can be performed on the participants to increase the sensitivity in detecting POI. For example, in another embodiment, skew analysis (third moment) and kurtosis (fourth moment) of the delay profile can be used to determine that a POI is occurring for some vehicles along the route. In looking at the skew analysis, we look for an increased forward component than that of the overall random population of vehicles within the bundle. The forward moment indicates that a small subset of participants in the route are stopping longer than is typical. Similarly, there is a likely POI if the kurtosis is flatter (platykuric, having a wide and generally flat peak around the mean), thereby not having sudden peaks, for the vehicles leaving airport than for the overall random population of vehicles (the control group). The likelihood of a POI, thus, can be calculated by multiplying the likelihood values derived from a single statistical model using mean and standard deviation, as well as the additional values obtained from analyzing the skew and kurtosis.
In accordance with another aspect of the invention, to further pin point POIs, the statistical analysis can be performed at different times to detect patterns of behavior that occur during different times. For example, the probe data can be obtained during different times of the day, different times of the week, different times of the month, or during different times of the year. In the case of the airport example, the vehicle traffic is typically greater during times later in the day, and thus, may not correspond with rush hour traffic which exhibits a different profile. In these cases, the time of day characteristics of the selected control group should be extended to the general population to be compared. This can be done by comparing the group behavior against that of the randomized subset of the general population, selected to have the same time of day statistical profile.
In accordance with another aspect of the invention, different analysis techniques can be used to interpret the data. For example, a profile of specific stops at a location that exceed a duration threshold of a predetermined period of time, such as 2 minutes, can be generated. As shown in FIG. 5, the airport bundle 9 in 5B exhibits delays in time distinct from the intersection 24, thereby further evidencing and corroborating information that there is POI, and in addition, identifying the location(s) of the POI within the route. For example, as with the evidence in FIG. 4D, column L indicating a slowdown at 65 meters prior to the intersection 24, the delay profile detected via probe data indicates a spike 65 meters prior to the intersection 24. Accordingly, the location of the original maneuver and intersection can now be used to position an entrance (E) to the POI within the street network 18.
The entrance E to the POI can be added automatically to the database, or it can be added manually upon verification, such as via aerial photography, satellite imagery, business and social networking websites, or city plans and maps, for example. Manual editing may be used in naming and deriving type or other information for the POI. In naming and deriving the type or other information for the POI, heuristics based on travel time and behavior, for example, can imply a POI type. Once the POI location has identified, a subset of traces within the selected group are selected which exhibit uncharacteristic delays compared to the overall control population for the particular maneuver. These uncharacteristic delays are then analyzed for time of day, time of week, etc. Any number of heuristic rules based on the culture and customs of the area can be applied. For example, certain areas may exhibit different socially accepted times for various meals (e.g. delay at such characteristic times may indicate restaurant), for worship (e.g. delay at such times may indicate a place of worship), for hotel check-in, etc. Other heuristics could indicate a window of time during which a POI is operational, wherein the delays that deemed to be arrival times may be analyzed and compared for week days versus weekends, thus indicating different times of operation, such as M-F 8:00am to 7:00pm, Saturday 8:00am to 5:00pm, and closed Sunday, for example. In addition, heuristics can be used to compare similar types of POI, such as hotels, for example, to indicate certain hotels as being preferred over other hotels based on frequency of occurrences.
Each of the aforementioned pieces of information obtained can be automatically attributed to the entrance point of the POI, or they can be slated for manual entering upon further investigation. In the case of frequency of POI visits, the information can be used to prioritize the manual research for verification purposes—that is, new POI locations receiving the most delays that are deemed to be arrivals can be given highest priority for study. Accordingly, a POI database of most frequently visited sites can be corroborated first.
It should be recognized that the airport example discussed above can be applied to virtually any scenario, particularly those locations having a well defined exit route, wherein vehicles leaving the location can be differentiated from a general population.
Obviously, many modifications and variations of the present invention are possible in light of the above teachings. It is, therefore, to be understood that within the scope of the appended claims, the invention may be practiced otherwise than as specifically described.

Claims

1. A method of analyzing points of interest using traces from probe data, comprising:

providing a database of a digital vector map configured to store a plurality of traces representing roads of a navigable street network;

collecting probe data from vehicles traveling along said traces;

bundling a group of select traces having routes with a common origin and at least one divergence point downstream from said origin;

building a database of vehicle maneuvers over said routes;

computing at least one of average speeds, delay time, or delay profile, of said group of select traces of vehicles traversing said vehicle maneuvers;

computing at least one of average speeds, delay time, or delay profile, of all vehicles traversing said vehicle maneuvers; comparing the at least one of average speeds, delay time, or delay profile from the said group of select traces of vehicles with the at least one of average speeds, delay time, or delay profile from all vehicles traversing said vehicle maneuvers; and

determining, from said comparison, information associated with one or more points of interest along said routes.

2. The method of claim 1 in which said determining information associated with one or more points of interest along said routes comprises:

determining at least one of the location, type of establishment, or hours of operation, of a point of interest along said routes.

3. The method of claim 1 further including calculating the skew of the delay times.

4. The method of claim 1 further including calculating the kurtosis of the delay times.

5. The method of claim 1 further including performing the computing steps during at least one of a predetermined time of day, week, month and year.

6. The method of claim 1 further including generating a profile of specific locations along the routes of the delay times.

7. A method of analyzing points of interest using traces from probe data, comprising:

providing a database of a digital vector map configured to store a plurality of traces representing a navigable network;

collecting probe data from travelers traveling along said traces;

building a database of maneuvers over said routes;

computing at least one of average speeds, delay time, or delay profile, of said group of select traces of travelers traversing said maneuvers;

computing at least one of average speeds, delay time, or delay profile, of all travelers traversing said maneuvers;

comparing the at least one of average speeds, delay time, or delay profile from the said group of select traces of travelers with the at least one of average speeds, delay time, or delay profile from all travelers traversing said maneuvers; and

determining, from said comparison, at least one of the location, type of establishment, or hours of operation, of a point of interest along said routes.