Toward a crowdsourcing solution to identify high-risk highway segments through mining driving jerks

Area of Science:

• Road safety
• Transportation engineering
• Data science

Background:

• Traffic crashes are a major global cause of preventable deaths.
• Identifying high-risk road segments is crucial for safety specialists and road users.
• Existing methods may not proactively identify all hazardous road segments.

Purpose of the Study:

• To develop and evaluate a novel crowdsourcing solution for identifying high-risk highway segments.
• To utilize driving jerks as a surrogate measure for traffic risk assessment.
• To compare the effectiveness of the proposed method against crash-rate data.

Main Methods:

• Collected naturalistic driving data to calculate driving jerks.
• Identified "unsafe" drivers based on their jerk-ratio.
• Applied an improved line-constrained clustering method to identify individual jerk clusters on road segments.
• Overlapped individual jerk clusters with road networks to detect potential risky segments.
• Synthesized results from multiple participants to determine final high-risk segments.

Main Results:

• The crowdsourcing solution effectively identified high-risk road segments.
• The method achieved an estimated accuracy of 75% in detecting hazardous segments.
• Driving jerks proved to be a valuable surrogate measure for identifying road risks.

Conclusions:

• The proposed crowdsourcing solution accurately identifies high-risk road segments using driving jerk analysis.
• This method allows for the proactive identification of hazardous road segments before crashes occur.
• Safety specialists can leverage this approach for enhanced road safety management.