Dynamic Driver Scoring Using Event-Based Data Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Usage-based insurance (UBI) requires significant data collection, leading to excessive data usage, and typical driving scores do not provide real-time benefits for drivers, making it challenging for UBI to coexist with other data-based services and updates.
Innovation Solution
A method is developed to generate a dynamic driver score in real-time using data analytics, which collects data over two months, assigns frames based on acceleration and speed thresholds, determines adverse behaviors using machine learning, and provides notifications to drivers for improved safety, thereby reducing data usage to approximately 7 MB or less per month.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If one-second period data is collected from a vehicle for UBI, then driving score calculation is enabled, but data usage becomes excessive (35 MB per month)
Solution Approach 1:
The patent segments continuous one-second data into selective event-based data points. Instead of collecting all one-second intervals, the system segments data collection to focus only on relevant driving events (harsh braking, harsh acceleration, rapid deceleration) identified through machine learning analysis, thereby reducing overall data volume while maintaining scoring accuracy
Solution Approach 2:
The patent extracts only the essential features needed for driver scoring from the complete one-second data stream. By using machine learning to identify and extract specific adverse behavior events rather than processing all continuous data, the system achieves accurate driving scores with significantly reduced data collection (from 35 MB to 7 MB per month)
2Loss of information
If typical driving scores are calculated, then insurance rate determination is enabled, but real-time safety benefit for drivers is not provided
Solution Approach 1:
The patent implements real-time feedback by immediately notifying drivers when adverse behavior is detected through machine learning analysis. Instead of providing only periodic scoring, the system continuously monitors driving behavior and provides instant feedback notifications, enabling drivers to correct behaviors in real-time and improving both safety outcomes and driver engagement
Solution Approach 2:
The patent uses machine learning to predict and identify adverse behaviors as they are occurring or about to occur, allowing the system to provide preventive warnings to drivers before harmful events fully develop. This preliminary detection and warning capability enables proactive safety interventions rather than reactive scoring alone
Data Source
AI summary
A method and apparatus may be used in a vehicle to generate a dynamic driver score. The method may include collecting data. The data may be collected at a predetermined interval. The data may include one or more frames. Each frame may correspond to the predetermined interval and may be divided into segments. Each frame may include a first acceleration data and a second acceleration data. The first acceleration data may be associated with an acceleration of a vehicle at a first segment of the frame, and the second acceleration data may be associated with an acceleration of the vehicle at a second segment of the frame. The frames may be grouped using machine learning methods to determine the driver score. The driver score may be used to modify driving behavior.