Authenticating Drivers

Abstract

One or more devices in a data analysis computing system may be configured to receive and analyze movement data and determine driving trips based on the received data. The driving trips may be used along with the movement data to authenticate drivers based on a determined driver profile.

What is claimed is: 
     
         1 . A method comprising:
 receiving, by a computing device, vehicle sensor data collected by one or more sensors during a driving trip;   identifying, by the computing device and in a training set of driver trip data, known points of interest (POI) clusters;   determining, by the computing device and based on comparing the vehicle sensor data to the training set, a probability that endpoints of the vehicle sensor data belong to the known POI clusters;   determining, by the computing device and based on the probability that the endpoints of the vehicle sensor data belong to the known POI clusters, whether the vehicle sensor data represents a routine driving trip;   selecting, by the computing device and based on whether the vehicle sensor data represents the routine driving trip, one of a plurality of machine learning models to use to identify a driver associated with the driving trip;   determining, by the computing device, from the vehicle sensor data and using the selected machine learning model, the driver associated with the driving trip; and   displaying, by the computing device, information corresponding to the determined driver.   
     
     
         2 . The method of  claim 1 , further comprising:
 determining, based on the probability satisfying a threshold value, that the vehicle sensor data represents the routine driving trip,   wherein the selected machine learning model comprises a first machine learning model, of the plurality of machine learning models, and   wherein the first machine learning model uses routine-based features extracted from the vehicle sensor data to determine the driver associated with the driving trip.   
     
     
         3 . The method of  claim 1 , further comprising:
 causing, based on determining that the vehicle sensor data represents the routine driving trip, the selected machine learning model to perform:
 analyzing the vehicle sensor data to determine a driving pattern associated with the driving trip; 
 determining, based on the driving pattern, a driving profile for the driving trip; and 
 determining, based on comparing the driving profile to previously determined driver profiles and a background driver profile, the driver associated with the driving trip. 
   
     
     
         4 . The method of  claim 3 , wherein analyzing the vehicle sensor data to determine the driving pattern comprises:
 analyzing the vehicle sensor data to determine one or more of:
 a stopping point corresponding to a location at which the vehicle stopped during the driving trip, or 
 a total number of turns during the driving trip; and 
   determining the driving pattern associated with the driving trip, based at least in part on the stopping point or the total number of turns.   
     
     
         5 . The method of  claim 1 , further comprising:
 determining, based on the probability not satisfying a threshold value, that the vehicle sensor data represents an anomalous driving trip,   wherein the selected machine learning model comprises a second machine learning model, of the plurality of machine learning models, and   wherein the second machine learning model uses control-based features extracted from the vehicle sensor data to determine the driver associated with the driving trip.   
     
     
         6 . The method of  claim 1 , further comprising:
 causing, based on determining that the vehicle sensor data represents an anomalous driving trip, the selected machine learning model to perform:
 analyzing the vehicle sensor data to determine a plurality of time-series data; 
 dividing each of the plurality of time-series data into a plurality of overlapping window frames of a predetermined length; 
 for each of the plurality of overlapping window frames, analyzing corresponding data for the window frame to determine a first probability score reflecting a probability that the window frame comprises driver trip data and a second probability score reflecting a probability that the window frame comprises non-driver trip data; 
 calculating, using the first probability scores for each of the plurality of overlapping window frames, a first total probability score reflecting a probability that the driving trip comprises driver-trip data; 
 calculating, using the second probability scores for each of the plurality of overlapping window frames, a second total probability score reflecting a probability that the driving trip comprises non-driver trip data; and 
 determining, based on a comparison of the first total probability score and the second total probability score, the driver associated with the driving trip. 
   
     
     
         7 . The method of  claim 6 , wherein the plurality of time-series data comprises control-based features corresponding to at least one of: a steering wheel angle, a gas pedal position, a brake pedal position, or one or more physical parameters, and
 wherein the one or more physical parameters comprise at least one of: speed, course, acceleration, jerk, angular speed, angular acceleration, angular jerk, or power per mass ratio.   
     
     
         8 . The method of  claim 6 , wherein analyzing the corresponding data for the window frame to determine the first probability score and the second probability score comprises:
 generating, using control-based features extracted from the corresponding data for the window frame, a feature vector for the window frame; and   determining a trip state associated with the feature vector for the window frame, and   wherein the method further comprises:
 retrieving a driver probability profile associated with a user of the computing device; 
 generating, based on comparing the feature vector for the window frame to a portion of the driver probability profile determined to correspond to the determined trip state associated with the feature vector for the window frame, the first probability score; 
 retrieving an average driver probability profile associated with an average driver; and 
 generating, based on comparing the feature vector for the window frame to a portion of the average driver probability profile determined to correspond to the determined trip state associated with the feature vector for the window frame, the second probability score. 
   
     
     
         9 . A computing device comprising:
 a processor; and   memory storing computer-executable instructions that, when executed by the processor, cause the computing device to:
 receive vehicle sensor data collected by one or more sensors during a driving trip; 
 identify, in a training set of driver trip data, known points of interest (POI) clusters; 
 determine, based on comparing the vehicle sensor data to the training set, a probability that endpoints of the vehicle sensor data belong to the known POI clusters; 
 determine, based on the probability that the endpoints of the vehicle sensor data belong to the known POI clusters, whether the vehicle sensor data represents a routine driving trip; 
 select, based on whether the vehicle sensor data represents the routine driving trip, one of a plurality of machine learning models to use to identify a driver associated with the driving trip; 
 determine, from the vehicle sensor data and using the selected machine learning model, the driver associated with the driving trip; and 
 display information corresponding to the determined driver. 
   
     
     
         10 . The computing device of  claim 9 , wherein the instructions, when executed by the processor, further cause the computing device to:
 determine, based on the probability satisfying a threshold value, that the vehicle sensor data represents the routine driving trip,   wherein the selected machine learning model comprises a first machine learning model, of the plurality of machine learning models, and   wherein the first machine learning model uses routine-based features extracted from the vehicle sensor data to determine the driver associated with the driving trip.   
     
     
         11 . The computing device of  claim 9 , wherein the instructions, when executed by the processor, further cause the computing device to:
 cause, based on determining that the vehicle sensor data represents the routine driving trip, the selected machine learning model to:
 analyze the vehicle sensor data to determine one or more of:
 a stopping point corresponding to a location at which the vehicle stopped during the driving trip, or 
 a total number of turns during the driving trip; and 
 
 determine, based at least in part on the stopping point or the total number of turns, a driving pattern associated with the driving trip; 
 determine, based on the driving pattern, a driving profile for the driving trip; and 
 determine, based on comparing the driving profile to previously determined driver profiles and a background driver profile, the driver associated with the driving trip. 
   
     
     
         12 . The computing device of  claim 9 , wherein the instructions, when executed by the processor, further cause the computing device to:
 determine, based on the probability not satisfying a threshold value, that the vehicle sensor data represents an anomalous driving trip,   wherein the selected machine learning model comprises a second machine learning model, of the plurality of machine learning models, and   wherein the second machine learning model uses control-based features extracted from the vehicle sensor data to determine the driver associated with the driving trip.   
     
     
         13 . The computing device of  claim 9 , wherein the instructions, when executed by the processor, further cause the computing device to:
 cause, based on determining that the vehicle sensor data represents an anomalous driving trip, the selected machine learning model to:
 analyze the vehicle sensor data to determine a plurality of time-series data; 
 device each of the plurality of time-series data into a plurality of overlapping window frames of a predetermined length; 
 for each of the plurality of overlapping window frames, analyze corresponding data for the window frame to determine a first probability score reflecting a probability that the window frame comprises driver trip data and a second probability score reflecting a probability that the window frame comprises non-driver trip data; 
 calculate, using the first probability scores for each of the plurality of overlapping window frames, a first total probability score reflecting a probability that the driving trip comprises driver-trip data; 
 calculate, using the second probability scores for each of the plurality of overlapping window frames, a second total probability score reflecting a probability that the driving trip comprises non-driver trip data; and 
 determine, based on a comparison of the first total probability score and the second total probability score, the driver associated with the driving trip. 
   
     
     
         14 . The computing device of  claim 13 , wherein the instructions, when executed by the processor, cause the computing device to analyze the corresponding data for the window frame to determine the first probability score and the second probability score by:
 generating, using control-based features extracted from the corresponding data for the window frame, a feature vector for the window frame; and   determining a trip state associated with the feature vector for the window frame, and   wherein the instructions, when executed by the processor, further cause the computing device to:
 retrieve a driver probability profile associated with a user of the computing device; 
 generate, based on comparing the feature vector for the window frame to a portion of the driver probability profile determined to correspond to the determined trip state associated with the feature vector for the window frame, the first probability score; 
 retrieve an average driver probability profile associated with an average driver; and 
 generate, based on comparing the feature vector for the window frame to a portion of the average driver probability profile determined to correspond to the determined trip state associated with the feature vector for the window frame, the second probability score. 
   
     
     
         15 . A non-transitory, computer-readable storage medium storing instructions that, when executed by a processor of a computing device, cause the computing device to:
 receive vehicle sensor data collected by one or more sensors during a driving trip;   identify, in a training set of driver trip data, known points of interest (POI) clusters;   determine, based on comparing the vehicle sensor data to the training set, a probability that endpoints of the vehicle sensor data belong to the known POI clusters;   determine, based on the probability that the endpoints of the vehicle sensor data belong to the known POI clusters, whether the vehicle sensor data represents a routine driving trip;   select, based on whether the vehicle sensor data represents the routine driving trip, one of a plurality of machine learning models to use to identify a driver associated with the driving trip;   determine, from the vehicle sensor data and using the selected machine learning model, the driver associated with the driving trip; and   display information corresponding to the determined driver.   
     
     
         16 . The non-transitory, computer-readable storage medium of  claim 15 , wherein the instructions, when executed by the processor, further cause the computing device to:
 determine, based on the probability satisfying a threshold value, that the vehicle sensor data represents the routine driving trip,   wherein the selected machine learning model comprises a first machine learning model, of the plurality of machine learning models, and   wherein the first machine learning model uses routine-based features extracted from the vehicle sensor data to determine the driver associated with the driving trip.   
     
     
         17 . The non-transitory, computer-readable storage medium of  claim 15 , wherein the instructions, when executed by the processor, further cause the computing device to:
 cause, based on determining that the vehicle sensor data represents the routine driving trip, the selected machine learning model to:
 analyze the vehicle sensor data to determine one or more of:
 a stopping point corresponding to a location at which the vehicle stopped during the driving trip, or 
 a total number of turns during the driving trip; and 
 
 determine, based at least in part on the stopping point or the total number of turns, a driving pattern associated with the driving trip; 
 determine, based on the driving pattern, a driving profile for the driving trip; and 
 determine, based on comparing the driving profile to previously determined driver profiles and a background driver profile, the driver associated with the driving trip. 
   
     
     
         18 . The non-transitory, computer-readable storage medium of  claim 15 , wherein the instructions, when executed by the processor, further cause the computing device to:
 determine, based on the probability not satisfying a threshold value, that the vehicle sensor data represents an anomalous driving trip,   wherein the selected machine learning model comprises a second machine learning model, of the plurality of machine learning models, and   wherein the second machine learning model uses control-based features extracted from the vehicle sensor data to determine the driver associated with the driving trip.   
     
     
         19 . The non-transitory, computer-readable storage medium of  claim 15 , wherein the instructions, when executed by the processor, further cause the computing device to:
 cause, based on determining that the vehicle sensor data represents an anomalous driving trip, the selected machine learning model to:
 analyze the vehicle sensor data to determine a plurality of time-series data; 
 device each of the plurality of time-series data into a plurality of overlapping window frames of a predetermined length; 
 for each of the plurality of overlapping window frames, analyze corresponding data for the window frame to determine a first probability score reflecting a probability that the window frame comprises driver trip data and a second probability score reflecting a probability that the window frame comprises non-driver trip data; 
 calculate, using the first probability scores for each of the plurality of overlapping window frames, a first total probability score reflecting a probability that the driving trip comprises driver-trip data; 
 calculate, using the second probability scores for each of the plurality of overlapping window frames, a second total probability score reflecting a probability that the driving trip comprises non-driver trip data; and 
 determine, based on a comparison of the first total probability score and the second total probability score, the driver associated with the driving trip. 
   
     
     
         20 . The non-transitory, computer-readable storage medium of  claim 15 , wherein the instructions, when executed by the processor, cause the computing device to analyze the corresponding data for the window frame to determine the first probability score and the second probability score by:
 generating, using control-based features extracted from the corresponding data for the window frame, a feature vector for the window frame; and   determining a trip state associated with the feature vector for the window frame, and   wherein the instructions, when executed by the processor, further cause the computing device to:
 retrieve a driver probability profile associated with a user of the computing device; 
 generate, based on comparing the feature vector for the window frame to a portion of the driver probability profile determined to correspond to the determined trip state associated with the feature vector for the window frame, the first probability score; 
 retrieve an average driver probability profile associated with an average driver; and 
 generate, based on comparing the feature vector for the window frame to a portion of the average driver probability profile determined to correspond to the determined trip state associated with the feature vector for the window frame, the second probability score.