techlist.io
KR EN

traffic-analysis

2 posts

google

Hard-braking events as indicators of road segment crash risk (opens in new tab)

Google Research has established a statistically significant correlation between hard-braking events (HBEs) collected via Android Auto and actual road crash rates. By utilizing HBEs as a "leading" indicator rather than relying on sparse, lagging historical crash data, researchers can proactively identify high-risk road segments with much greater speed and spatial granularity. This validation suggests that connected vehicle data can serve as a scalable proxy for traditional safety assessments. ### Data Density and Scalability * HBEs—defined as forward deceleration exceeding -3m/s²—provide a signal that is 18 times denser than reported crash data. * While crashes are statistically rare and can take years to provide a valid safety profile for a specific road segment, HBEs offer a continuous stream of information. * This high density allows for the creation of a comprehensive "safety map" that includes local and arterial roads where crash reporting is often inconsistent or sparse. ### Statistical Validation of HBEs * Researchers employed negative binomial regression models to analyze 10 years of public crash data from California and Virginia alongside anonymized HBE data. * The models controlled for confounding factors such as traffic volume, segment length, road type (local, arterial, highway), and infrastructure dynamics like slope and lane changes. * The results confirmed a consistent positive association between HBE frequency and crash rates across all road types, proving HBEs are a reliable surrogate for risk regardless of geography. ### High-Risk Identification Case Study * An analysis of a freeway merge connecting Highway 101 and Highway 880 in California served as a practical validation of the metric. * This specific segment was found to have an HBE rate 70 times higher than the state average, correlating with a historical record of one crash every six weeks. * The HBE signal successfully flagged this location as being in the top 1% of high-risk segments without needing years of collision reports to confirm the danger, demonstrating its utility in identifying "black spots" early. ### Real-World Application and Road Management * Validating HBEs transforms raw sensor data into a trusted tool for urban planners and road authorities to perform network-wide safety assessments. * This approach allows for proactive infrastructure interventions, such as adjusting signage or merge patterns, before fatalities or injuries occur. * The findings support the integration of connected vehicle insights into platforms like Google Maps to help authorities manage road safety more dynamically.

traffic-analysispredictive-modelingandroid-autoconnected-vehicle-data+3
google

How we created HOV-specific ETAs in Google Maps (opens in new tab)

Google Maps has enhanced its routing capabilities by introducing HOV-specific ETAs, addressing the significant speed differences between carpool and general lanes. This was achieved through a novel unsupervised learning approach that classifies historical trips into HOV or non-HOV categories without initial manual labels. The resulting system enables more precise travel predictions, helping users optimize their commutes and supporting the shift toward sustainable travel modes. ### Segment-Level Speed Distribution * The model analyzes trip segments within short, 15-minute time windows to identify patterns in aggregated, anonymized traffic data. * During peak traffic hours, researchers often observe a bimodal speed distribution where HOV lanes maintain significantly higher average speeds compared to general lanes. * The classification system distinguishes between "Scenario A," where the speed gap is dramatic (e.g., 65 mph vs. 25 mph), and "Scenario B," where HOV lanes are only marginally faster, ensuring accurate modeling even when benefits are minimal. * Individual trip points, including speed and observation time, are processed collectively to determine if a specific segment of a journey occurred in a restricted lane. ### Incorporating Lateral Distance and Soft Clustering * To refine accuracy beyond simple speed metrics, the model incorporates the estimated lateral distance of a vehicle from the center of the road. * While GPS data is inherently noisy, this spatial information helps identify lane-specific behaviors by mapping trip points to the known physical location of HOV lanes (e.g., the far-left lanes). * The system employs soft clustering techniques, calculating the probability of a point belonging to a specific cluster rather than using hard binary assignments, which better manages borderline data points. * Temporal clustering via a weighted median approach is used to prioritize more recent traffic observations, ensuring the model accounts for the most current road conditions and availability constraints. By integrating these segment-level classifications into full-trip analyses, Google Maps can train its ETA prediction models on high-fidelity, lane-specific data. This implementation provides users with a more realistic view of their travel options, encouraging the use of high-occupancy lanes to reduce individual travel time, urban congestion, and overall emissions.

traffic-analysisaimachine-learningeta-prediction+5