Systems and methods for generating vehicle safety scores and predicting vehicle collision probabilities

Patent metadata

Field	Value
Publication number	US-2025035457-A1
Application number	US-202418885909-A
Country	US
Kind code	A1
Filing date	Sep 16, 2024
Priority date	Nov 15, 2022
Publication date	Jan 30, 2025
Grant date	—

Abstract

Systems and methods for predicting collision probabilities are provided. The methods involve operating at least one processor to: retrieve vehicle data originating from a telematics device installed in a vehicle, the vehicle data including location data and a plurality of safety exception events performed by the vehicle, the plurality of safety exception events including a plurality of exception event types; identify a plurality of road network edges traveled by the vehicle based on the location data; determine an aggregated area collision rate based on the plurality of road network edges; determine a plurality of exception rates based on the vehicle data, each exception rate representing a normalized rate of occurrence of one of the exception event types; and determine a collision probability using at least one machine learning model on the plurality of exception rates and the aggregated area collision rate, the collision probability representing a risk of collision.

First claim

1 . A system for predicting collision probabilities, the system comprising:
at least one data store operable to store vehicle data originating from a telematics device installed in a vehicle, the vehicle data comprising location data and a plurality of safety exception events performed by the vehicle, the plurality of safety exception events comprising a plurality of exception event types; at least one processor in communication with the at least one data store, the at least one processor operable to:
retrieve the vehicle data;
identify a plurality of road network edges traveled by the vehicle based on the location data;
determine an aggregated area collision rate based on the plurality of road network edges;
determine a plurality of exception rates based on the vehicle data, each exception rate representing a normalized rate of occurrence of one of the exception event types; and
determine a collision probability using at least one machine learning model on the plurality of exception rates and the aggregated area collision rate, the collision probability representing a risk of collision.

2 . The system of claim 1 , wherein the at least one machine learning model comprises a decision tree.

3 . The system of claim 1 , wherein the at least one processor is operable to:
retrieve a plurality of predetermined area collision rates, a predetermined area collision rate retrieved for each road network edge; and determine the aggregated area collision rate based on the plurality of predetermined area collision rates.

4 . The system of claim 1 , wherein the at least one processor is operable to:
determine a vehicle collision probability benchmark for the vehicle based on the collision probability and a plurality of collision probabilities for a plurality of comparable vehicles.

5 . The system of claim 4 , wherein the at least one processor is operable to:
identify the plurality of comparable vehicles from a plurality of vehicles by:
using a clustering algorithm to identify a plurality of first vehicle clusters based on an area of operation of each vehicle in the plurality of vehicles; and
using a comparison algorithm to identify a plurality of second vehicle clusters based on the centroid of each first vehicle cluster; and
identify the second vehicle cluster containing the vehicle as the plurality of comparable vehicles.

6 . The system of claim 1 , wherein the at least one processor is operable to:
determine a fleet collision probability for a fleet comprising the vehicle based on the collision probability and a vehicle collision probability for each other vehicle in the fleet.

7 . The system of claim 6 , wherein the at least one processor is operable to:
determine a fleet collision probability benchmark for the fleet based on the fleet collision probability for the fleet and a plurality of fleet collision probabilities for a plurality of comparable fleets.

8 . The system of claim 7 , wherein the at least one processor is operable to:
identify the plurality of comparable fleets from a plurality of fleets by using a clustering algorithm to identify a plurality of fleet clusters based on a vehicle type of each vehicle in each fleet in the plurality of fleets; and identify the fleet cluster containing the fleet as the plurality of comparable fleets.

9 . The system of claim 1 , wherein:
the plurality of exception event types comprises harsh events and speeding events; and the plurality of exception rates comprises harsh event rates and speeding event rates.

10 . A method for predicting collision probabilities, the method comprising operating at least one processor to:
retrieve vehicle data originating from a telematics device installed in a vehicle, the vehicle data comprising location data and a plurality of safety exception events performed by the vehicle, the plurality of safety exception events comprising a plurality of exception event types; identify a plurality of road network edges traveled by the vehicle based on the location data; determine an aggregated area collision rate based on the plurality of road network edges; determine a plurality of exception rates based on the vehicle data, each exception rate representing a normalized rate of occurrence of one of the exception event types; and determine a collision probability using at least one machine learning model on the plurality of exception rates and the aggregated area collision rate, the collision probability representing a risk of collision.

11 . The method of claim 10 , wherein the at least one machine learning model comprises a decision tree.

12 . The method of claim 10 , further comprising operating the at least one processor to:
retrieve a plurality of predetermined area collision rates, a predetermined area collision rate retrieved for each road network edge; and determine the aggregated area collision rate based on the plurality of predetermined area collision rates.

13 . The method of claim 10 , further comprising operating the at least one processor to:
determine a vehicle collision probability benchmark for the vehicle based on the collision probability and a plurality of collision probabilities for a plurality of comparable vehicles.

14 . The method of claim 13 , further comprising operating the at least one processor to:
identify the plurality of comparable vehicles from a plurality of vehicles by:
using a clustering algorithm to identify a plurality of first vehicle clusters based on an area of operation of each vehicle in the plurality of vehicles; and
using a comparison algorithm to identify a plurality of second vehicle clusters based on the centroid of each first vehicle cluster; and
identify the second vehicle cluster containing the vehicle as the plurality of comparable vehicles.

15 . The method of claim 10 , further comprising operating the at least one processor to:
determine a fleet collision probability for a fleet comprising the vehicle based on the collision probability and a vehicle collision probability for each other vehicle in the fleet.

16 . The method of claim 15 , further comprising operating the at least one processor to:
determine a fleet collision probability benchmark for the fleet based on the fleet collision probability for the fleet and a plurality of fleet collision probabilities for a plurality of comparable fleets.

17 . The method of claim 16 , further comprising operating the at least one processor to:
identify the plurality of comparable fleets from a plurality of fleets by using a clustering algorithm to identify a plurality of fleet clusters based on a vehicle type of each vehicle in each fleet in the plurality of fleets; and identify the fleet cluster containing the fleet as the plurality of comparable fleets.

18 . The method of claim 10 , wherein:
the plurality of exception event types comprises harsh events and speeding events; and the plurality of exception rates comprises harsh event rates and speeding event rates.

19 . A non-transitory computer readable medium having instructions stored thereon executable by at least one processor to implement a method for predicting collision probabilities, the method comprising operating the at least one processor to:
retrieve vehicle data originating from a telematics device installed in a vehicle, the vehicle data comprising location data and a plurality of safety exception events performed by the vehicle, the plurality of safety exception events comprising a plurality of exception event types; identify a plural

Assignees

• Geotab Inc

Inventors

• Zhang Xin
• Hines Gregory Gordon Douglas
• Yu Jiawei
• Petersen Willem
• Tiwari Tuhin
• Murugesan Meenakshi Sundaram
• Siddique Javed
• Yan Jason Jiajie
• Durgam Narasimha Rao
• Zhang Li
• Manjunath Vinay Kiran
• Zhou Xinrong
• Chen Yujie
• Xu Chenyue
• Vazquez Luis Perez

Classifications

• B60W30/095

  Predicting travel path or likelihood of collision · CPC title

• G06Q10/04

  Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem" (market predictions or forecasting for commercial activities G06Q30/0202) · CPC title

• G06N20/00

  Machine learning · CPC title

• H04W4/029

  Location-based management or tracking services · CPC title

• G06N20/20 Primary

  Ensemble learning · CPC title