Driver Scoring Using Alert Clustering for Road Bias Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional driver scoring techniques fail to account for accident proneness of road segments, leading to biased driver score computations due to varying road conditions and infrastructure, which affects scoring accuracy.
Innovation Solution
A driver scoring system that determines road accident proneness (RAP) scores by clustering alerts and normalizing them based on criticality and trip frequency, integrating these scores into driver score calculations to reduce bias and enhance accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional driver scoring techniques are used, then the scoring process is simple, but the scoring accuracy is reduced due to accident proneness bias from varying road conditions
Solution Approach 1:
The patent segments the driver scoring process into multiple components: raw score calculation, road segment identification, alert clustering within segments, and bias adjustment. By dividing the scoring system into these distinct segments, it can apply specific processing (alert clustering) to remove accident proneness bias from road segments while keeping other parts of the system relatively simple.
Solution Approach 2:
The patent introduces an intermediary component (the alert clustering module that processes road segment data) between the raw driver behavior data and the final driver score. This intermediary calculates accident proneness scores for road segments and uses them to adjust individual alert scores, thereby eliminating bias without requiring complete system redesign.
2Measurement precision
If road type classification relies on lane markers and road infrastructure, then classification is straightforward, but accuracy is impaired in developing nations with poor infrastructure
Solution Approach 1:
The system uses alert data generated during normal vehicle operation to automatically classify road types and identify accident-prone segments. Instead of relying on external infrastructure data, the system serves itself by utilizing its own operational data (alerts, vehicle behavior) to improve road classification accuracy adaptively.
Solution Approach 2:
The patent changes the parameters used for road classification from static infrastructure features (lane markers) to dynamic operational parameters (alert frequency, alert types, vehicle behavior patterns). This allows the system to adapt to varying road conditions across different regions, including areas with poor infrastructure.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Techniques are disclosed to determine driver scoring that consider road accident proneness scores in driver score computations. The road accident proneness scores may be based on one or more alerts. Further, road types of roads may be determined based on alerts. The determination of the road type may use a machine learning model. In contrast to the conventional approaches, the use of road accident proneness scores in the computation of driver scores allows for the consideration of accident proneness of the road to improve upon conventional scoring techniques. Further, map data and generated maps may be improved based on determined road types.