This invention relates to a method of determining cost of insurance based on the correlation of information from actual accidents with the historical analysis of vehicle movement. A method is based on performing the actuarial calculation of the risk involved in insuring a vehicle, based on the calculation of the correlation of the data collected from the actual incidents of the subset of the vehicle sample fleet and driving characteristics obtained through historical pattern of the movement of a vehicle sample fleet equipped with GPS sensors for statistically significant period of time.

BACKGROUND OF THE INVENTION

The described invention relates to acquisition of vehicle movement data and its correlation with the actual accidents that happened within such controlled group of vehicles, in order to further refine the possibility of determining the proper premium of car insurance based on tracking data and correlation with actual accidents.

Alternative methods used for so called user based insurance premiums, collect data in the fashion similar to the methods described within this invention. Such data is generally focused on driver's behavior through collection of data on sudden acceleration, breaking, speeding or generally, commonly accepted signs of risky driving. Others collect data on position to determine the frequency of passages through dangerous segments of the roads or collect data on weather conditions and generate trigger events used for the scoring of drivers habits and common trips.

Although organizations like National Highway Safety Authority publishes statistics that show as the prevalent cause of accidents on American highways speeding, driving by night or crossing through intersections, generalization of this common mode failure mechanisms is not mathematically related to the historical behavior of drivers and vehicle trips and therefore cannot be successfully used in determination of the insurance premiums.

None of published methods have statistically founded calculation of risk based on the correlation of real accidents and collected data and therefore any influence on the cost of the premiums is arbitrary and based on some kind of heuristic that has no foundation in exact mathematical modeling. State of the Art

4533962 August 1985 Decker et al.

4608638 August 1986 Tsikos

4667336 May 1987 Best

4763745 August 1988 Eto et al.

4807179 February 1989 Clere et al.

4829434 May 1989 Karmel et al.

4843463 June 1989 Michetti

4843578 June 1989 Wade

4853720 August 1989 Onari et al.

4939652 July 1990 Steiner

4987541 January 1991 Levente et al.

4992943 February 1991 McCracken

5055851 October 1991 Sheffer

511 1289 May 1992 Lucas et al.

5189621 February 1993 Onari et al.

5223844 June 1993 Mansell et al.

5319374 June 1994 Desai et al.

5365451 November 1994 Wang et al.

5430432 July 1995 Camhi et al.

5499182 March 1996 Ousborne

5548273 August 1996 Nicol et al.

5550551 August 1996 Alesio

5638273 June 1997 Coiner et al.

5797134 August McMillan , et a

DESCRIPTION OF THE INVENTION

Invention described herein, shows the way of mathematical modeling that can clearly segregate higher risk driving (including the use of roads), as opposed to lower risk driving and therefore actuarial data arising from this invention has significant economic value for insurance companies. In accordance with the present invention, there is disclosed method of determining cost of insurance based on the correlation of information from actual accidents with the historical analysis of vehicle movement belonging to vehicle sample fleet and equipped with GPS sensors for statistically significant period of time. The resulting mathematical model is valid for the region where the GPS data has been collected.

By the term "region" in this invention is meant state, country, geographical region, city, county etc.

By the term "statistically significant period of time" in this invention is meant a period of time which should include the periodicity of patterns of data such as weather data, traffic congestions data and similar. It is preferable that such statistically significant period of time is 6 or more months. Most preferable is that the statistically significant period of time is one year or more.

The invention is based on a method of calculating probabilities of traffic accident in region, and comprises of the following steps:

Collection of data received from the vehicle onboard GPS tracking unit installed on the each vehicle of the vehicle sample fleet, during the statistically significant period of time;

Sampling and storing of the vehicle GPS data with the frequency of at least hundred meters of vehicle movement;

Using of a vehicle GPS data to precisely map match vehicle movement with the digital map and its corresponding road network during the statistically significant period of time;

Generation of driving characteristics of the vehicle sample fleet using map matching and post processing of the collected GPS data for vehicle sample fleet and calculation of distribution of resulting driving characteristics;

Collection of real accident data from the region during the statistically significant period of time, where the subset of vehicles from vehicle sample fleet is included;

Correlation of driving characteristics of the vehicle sample fleet with real accident data for the subset of vehicle sample fleet in the region in order to obtain correlation factors;

Use of obtained correlation factors in the actuarial calculation of the risks arising from such correlation and application of calculated risk on the determination of the increase or reduction of insurance premium for all the vehicles being equipped with GPS sensors for the statistically significant period of time.

The step of correlation of driving characteristics of the sample vehicles with real accident data includes further correlation of all said data with cross-referenced traffic data, weather report data, or driver data (if such data is collected) and any combination of the above including the combination of particular driving characteristics. For the purpose of collecting vehicle movement and position data, GPS tracking (telematics) unit are built into vehicles belonging to vehicle sample fleet ,thus having the ability to send data to server via GPRS or other wireless modem with the frequency dependent on vehicle speed so that every 100 meters a packet of data is send to server consisting of at minimum:

id of the vehicle

time stamp of the sample

gps coordinates as obtained by GPS sensor

instantaneous speed of vehicle, and such data packets are stored in the memory of the server for later analysis. ID of driver can be also added to above minimum data set in one of the embodiments of described invention

After described step, the raw GPS from the sensor is matched against the digital map data in order to digitally place the vehicle on the given road segment and glean from the map data the type of the road and speed limit superimposed on such particular road segment.

The obtained map matched vehicle speed is compared with the speed limit from the actual map data road segment in order to determine whether vehicle moves twenty or less kilometers (miles) per hour thus representing speeding event or with the speed that is twenty or more kilometers (miles) per hour over given road segment speed limit, thus representing significant speeding event.

The obtained number of speeding and significant speeding event is determined for the complete set of trips recorded for any particular vehicle for the statistically significant period of time in order to calculate the number of speeding and significant speeding events per kilometer (mile) driven for a given vehicle from the vehicle sample fleet.

After this step or in parallel with this step, all the stored records are analyzed with respect to their time stamp in order to determine whether trip occurred during the night or day. The total drive per night is then divided by the total time driven during the statistically significant period of time, for every vehicle from the vehicle sample fleet in order to determine the percentage of driving during the night within the total time driven.

After that or in parallel with the previous steps, the movement of the vehicle is analyzed in the context of road network geometry in order to determine the passage through or turning of the vehicle in the intersection with the analysis of the type of the crossing or turning with the respect to relationship of given road segments intersecting in terms of the road's right of way. In this fashion the number of right turns, left turns or crossings is counted separately for vehicles entering the intersection from the road with the right of way as opposed to those happening when the vehicle makes right, left turns or crosses the intersection from the road that yields to intersecting road. Result from this step is

the calculation of the various number of turns and crossings per kilometer driven obtained through detailed analysis for each vehicle from the vehicle sample fleet for the statistically significant period of time.

Furthermore, each map matched position of the vehicle is compared with the records from the database containing the location of the accidents that happened on given locations in order to calculate the total number of passing through such dangerous spots for each vehicle from the vehicle sample fleet in order to calculate the number of such passages per kilometer driven for the statistically significant period of time.

It is also preferred that every map matched record is used as the index to historical digital weather map in order to determine whether a trip happened during rainy, snowy or icy conditions on the road in order to calculate the percentage of the time driven in rainy, snowy or icy condition on the road. This percentages are calculated by dividing the total duration of the trips during rain, snow or ice condition on the road with the total duration of the trips for every vehicle from the vehicle sample fleet for the statistically significant period of time.

It is also preferred that every map matched record is compared with the speed profiles of the road segments in order to determine whether drive happened during rush hour in order to calculate the percentage of driving during rash hour which is calculated by dividing total time of rush hour trips with total time driven for every vehicle form the vehicle sample fleet for the statistically significant period of time.

Finally, every id of the vehicle from the vehicle sample fleet is used as the index to a database of all accidents to identify the subset of the vehicles from the vehicle sample fleet that were involved in the accident during the statistically significant period of time, in order to establish the time and the place as well as the circumstances of accident (age and sex of the driver as well as years of experience of a driver )

Those instances of accidents are correlated against the values of the driving characteristics described above as well as combination of those values and in combination with data describing circumstances of accidents.

The last step of the method is to use the obtained correlation for the actual calculation of risk of accident for every vehicle analyzed in order to calculate the reduction or increase in insurance premium. For this purpose the source of accident data for the subset of vehicle sample fleet should come from region's official traffic accident record keeper such as police reports or other sources of official traffic accident records for the region.

The driving characteristics of a vehicle sample fleet obtained through map matching and post processing of the collected GPS data for the vehicle sample fleet will consequently allow for the calculation of the distributions of driving characteristics that can include, but are not limited to:

1. speeding

2. driving through intersections

3. driving by night

4. driving during rush hours

5. driving in rainy, snowy and icy conditions

6. driving through dangerous road segments.

Data is collected and driving characteristics are calculated for all the vehicles from the vehicle sample fleet.

The size of the vehicle sample fleet must be large enough and uniformly represent all the vehicles participating in the traffic of a region. Furthermore, the size of the fleet must be large enough to exhibit the high confidence that subset of the vehicles from the vehicle sample fleet will be involved in the accidents during the period of the collection of data that is statistically significant. It is preferred that the size of the vehicle sample fleet is 1000 vehicles or more.

The actual data collected from such accidents for the subset of a vehicle sample fleet is than correlated with one or more driving characteristics derived from GPS tracking and map matching of the same GPS data, and thus produces the probabilities that one ore more driving characteristics may produce higher risk of accidents than average accident rate.

It is envisioned that in the further embodiment of this method calculated driving characteristics as well as accident data can be assigned to a particular person.

Calculated probabilities and their combination lead to exact calculation of insurance premiums that make insurance of the cars profitable depending on the size of the premium and the risk of the vehicles for every vehicle. Once when this model is derived for the vehicle sample fleet, it can be applied to all the vehicles from the region that participate in this type of use based insurance and allow equipping of the insured car with necessary GPS sensors and if desired with the identification of the driver.

BRIEF DESCRIPTION OF THE DRAWINGS

Figure 1 . - Flowchart describing acquisition data from the vehicle

Figure 2. - Flowchart describing calculation of correlation of derived GPS data and actual accident data in calculation of risk factors

DETAILED DESCRIPTION OF THE INVENTION

Referring now to the drawings, wherein the showings are for purposes of illustrating the preferred embodiments of the invention only and not for purposes of limiting the same, the Figures show an apparatus and method for calculating risk factors associated with the actual movement of the vehicles belonging to the fleet that is large enough and selected in a fashion to statistically significantly represents the overall set of the vehicles from some region or the country.

The method described herein is based on mathematically founded facts that derive actuarial data from the correlation of aforementioned driving characteristics and actual accidents with involvement of the vehicles form the same group that was used for the generation of such driving characteristics.

As shown in Fig. 1 each vehicle form the vehicle sample fleet, that statistically significantly represents all the vehicles from the country or the region, is equipped with the GPS sensor that collects the data continuously. Only when the vehicle moves, data packets containing GPS coordinates of the instantaneous position of the vehicle, vehicle ID, driver ID, vehicle ground speed and time when the sample was taken are sent to a server once when the position of vehicle changes for more than 100 m. Any other sampling that can provide basis for proper map matching post processing of such data can be used alternatively.

Collected data is send to a server and stored in the memory of the server for every vehicle, as it is generated in the field. Referring now to Fig 2. the data coming from GPS devices is used as the raw data that allows calculation of the driving characteristics used in the method described herein. In order to calculate those driving characteristics, the servers may have access to map data including traffic patterns and weather data for the region.

The calculated driving characteristics can be, but are not limited to: map matched positions of the vehicles ( vehicles glued to road segments) from the vehicle sample fleet

speeding event - every instance when ground speed of the vehicle from the vehicle sample fleet exceeds speed limit on the road segment identified through map matching of GPS data for that vehicle for up to 20 km/h. The number of such speeding events is calculated for every vehicle form the vehicle sample fleet for the statistically significant period of time

significant speeding event - every instance when ground speed of the vehicle from the vehicle sample fleet exceeds speed limit on the road segment identified through map matching of GPS data for that vehicle over 20 km/h. The number of such significant speeding events is calculated for every vehicle form the vehicle sample fleet for the statistically significant period of time percentage of driving during the night calculated using vehicle GPS position and the time of the day when the sample was taken. Percentage is calculated by identifying the total number of night driving samples and dividing them with total number of samples for every vehicle from the vehicle sample fleet. The percentage is calculated for every vehicle form the vehicle sample fleet for the statistically significant period of time number of crossings through intersections for every vehicle from the vehicle sample fleet coming from the road that yields to intersecting road, divided by total kilometers driven per each vehicle from the vehicle sample fleet. Such number of crossings is calculated for every vehicle form the vehicle sample fleet for the statistically significant period of time number of crossing through intersections for vehicles from the vehicle sample fleet coming from the road with the right of way to intersecting road, divided by total kilometers driven per each vehicle from the vehicle sample fleet. Such number of crossings is calculated for every vehicle form the vehicle sample fleet for the statistically significant period of time number of right turns on intersections for vehicles from the vehicle sample fleet making the turn from the road with the right of way, divided by total kilometers driven per each vehicle from the vehicle sample fleet. Such number of right turns is calculated for every vehicle form the vehicle sample fleet for the statistically significant period of time number of right turns on intersections for vehicles from the vehicle sample fleet making the turn from the road that yields to intersecting road, divided by total kilometers driven per each vehicle from the vehicle sample fleet. Such number of right turns is calculated for every vehicle form the vehicle sample fleet for the statistically significant period of time number of left turns on intersections for vehicles from the vehicle sample fleet making the turn from the road with the right of way, divided by total kilometers driven per each vehicle from the vehicle sample fleet. Such number of left turns is calculated for every vehicle form the vehicle sample fleet for the statistically significant period of time number of left turns on intersections for vehicles from the vehicle sample fleet making the turn from the road that yields to intersecting road, divided by total kilometers driven per each vehicle from the vehicle sample fleet. Such number of left turns is calculated for every vehicle form the vehicle sample fleet for the statistically significant period of time number of crossings through dangerous segments of the road divided by total kilometers driven for every vehicle from the vehicle sample fleet. Such number of crossings is calculated for every vehicle form the vehicle sample fleet for the statistically significant period of time time driven through rain, snow or ice by summing all the records for each vehicle from the vehicle sample fleet whose position used as a query in weather record database returned weather condition for such particular road segment and dividing it by total number of trips per vehicle from the vehicle sample fleet to obtain percentage of driving through rainy, snowy or icy conditions on the road. Such percentages are calculated for every vehicle form the vehicle sample fleet for the statistically significant period of time percentage of driving during rush hour calculated by dividing total duration of trips that, used as a query to speed profile database, returned rush hour condition divided by total number of trips per vehicle from the vehicle sample fleet. Such percentages are calculated for every vehicle form the vehicle sample fleet for the statistically significant period of time

Furthermore, still referring to Fig.2 , every vehicle Id is used as the query into Real Accidents Database that has records of all the accidents that happened in the analyzed region in the observed, statistically significant period of time to identify only those accidents where subset of vehicles from the vehicle sample fleet were involved. Now, it is possible to run correlation between above calculated driving characteristics and accidents involving subset of the vehicles from the vehicle sample fleet. The calculation of correlation happens on each driving characteristics and their combination to establish the strongest links between driving characteristics and accidents involving the subset of the vehicles form the vehicle sample fleet. For example , but not limited to only this example , any combination of parameters such as vehicles driving through intersection during the rainy, icy or snowy conditions and making the left turn from the road with the right of way may show strong correlation with the probability of accident happening . The calculation of that probability becomes de facto actuarial risk figure that can be applied to analysis of resulting premium. Furthermore, such actuarial figure can be applied to any vehicle participating in described use based insurance model assuming that such vehicles are equipped with GPS sensors, and not only to the vehicles form the vehicle sample fleet. The process of refining the model can be calibrated using described method after every year or statistically significant period of time in order to constantly address changing conditions in the traffic of particular country or the region. Furthermore, as mentioned above if the means for identifying the driver exists in the system the risk factors can be calculated for person driving the vehicle rather than a vehicle.

The invention has been described with reference to preferred embodiments. Obviously, modifications and alterations will occur to others upon a reading and understanding of the specification. It is our intention to include all such modifications and alterations insofar as they come within the scope of the appended claims or the equivalents thereof.