Driving Behavior Identification via Machine-Learned Key Figures
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for identifying driving behavior require external expert intervention, making it difficult to verify the significance of data and evaluate driving behavior accurately and efficiently, especially in large datasets, and do not provide transparent or objective assessments of accident risk.
Innovation Solution
A computer-implemented method that compresses driving data into key figures, uses machine-learned indicators to characterize driving behavior, and identifies risky maneuvers by combining sensor data with metadata, enabling automated and transparent evaluation of driving behavior and accident risk.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If automated methods are used to process large amounts of driving data, then productivity and processing speed are improved, but measurement precision and verification of data significance deteriorate
Solution Approach 1:
The patent introduces machine learning models as intermediaries between raw driving data and driving behavior assessment. These models process large volumes of sensor data automatically while providing statistically verifiable results through probability distributions and confidence intervals, thus maintaining both high processing speed and measurement precision
Solution Approach 2:
The patent replaces manual expert assessment (mechanical/heuristic system) with automated machine learning algorithms. This substitution enables processing of large datasets at high speed while maintaining objective, reproducible, and verifiable results through mathematical models rather than subjective expert judgment
2Measurement precision
If expert intervention is used to determine driving behavior profiles, then measurement precision is improved, but device complexity and processing time worsen
Solution Approach 1:
The patent enables the system to automatically learn and determine driving behavior patterns without requiring external expert intervention. The machine learning models self-adjust parameters and identify patterns from training data, providing accurate assessments while simplifying the system by removing the need for expert systems or manual configuration
Solution Approach 2:
The patent transforms the approach from fixed expert-defined rules to dynamic parameter learning. The system automatically adapts parameters such as acceleration thresholds, speed limits, and behavior weights based on training data, achieving accurate assessment without complex manual configuration or expert intervention
3Measurement precision
If comprehensive driving data is collected and processed, then measurement precision is improved, but loss of time and processing efficiency worsen
Solution Approach 1:
The patent performs preliminary processing by continuously collecting and pre-processing driving data in the background, preparing it for rapid analysis. This includes filtering, normalization, and feature extraction that can be performed quickly when assessment is needed, thus reducing actual processing time while maintaining comprehensive data analysis
Solution Approach 2:
The patent segments the driving data processing into distinct modules: sensor data collection, feature extraction, machine learning inference, and result generation. This segmentation allows parallel processing of different data streams and enables the system to process comprehensive data efficiently by handling different aspects simultaneously rather than sequentially
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A computer-implemented method for characterizing driving behavior, comprising the steps of: - obtaining driving data from a trip, wherein the driving data is assigned to at least one vehicle; - condensing the driving data into one or more key figures for a trip; - storing the condensed driving data; - determining at least one indicator for a driver's driving behavior, based on stored condensed driving data from multiple trips within a specific period; - outputting the at least one indicator. In an advantageous embodiment, the indicator is determined based on a machine-learned mapping.