Connected Car Data: The Next Generation of Surrogate Safety Analytics

[et_pb_section fb_built="1" admin_label="section" _builder_version="3.22" da_disable_devices="off|off|off" da_is_popup="off" da_exit_intent="off" da_has_close="on" da_alt_close="off" da_dark_close="off" da_not_modal="on" da_is_singular="off" da_with_loader="off" da_has_shadow="on"][et_pb_row admin_label="row" _builder_version="3.25" background_size="initial" background_position="top_left" background_repeat="repeat"][et_pb_column type="4_4" _builder_version="3.25" custom_padding="|||" custom_padding__hover="|||"][et_pb_text admin_label="Intro" _builder_version="4.7.3" background_size="initial" background_position="top_left" background_repeat="repeat"]

When an intersection sees a high frequency of hard braking, or a road section sees high volumes of vehicle speeding, one might think that it’s a high-risk area for road safety and proactive measures should be implemented to prevent crashes from occurring. Proving the intuition with data, however, hasn’t been easy.

Data such as hard braking or speeding – which was impossible to collect on scale a few years ago – are becoming available through connected car data and quickly being leveraged by innovative transportation agencies to identify safety “hot spots”, i.e., locations more prone to accidents. Such data can be collected for network-wide studies or for pinpointing a particular intersection or road segment, allowing road safety experts to investigate citizen complaints, prioritize projects, dig deeper into driver behaviours causing accidents, and design counter measures.

If your goals are to reduce crash frequency and crash severity, and to do it as early as possible, keep reading.

[/et_pb_text][/et_pb_column][/et_pb_row][et_pb_row _builder_version="4.7.3" _module_preset="default"][et_pb_column type="4_4" _builder_version="4.7.3" _module_preset="default"][et_pb_text admin_label="Rise of surrogate safety data" _builder_version="4.7.3" _module_preset="default" header_2_text_color="#0c71c3" header_3_text_align="left" header_3_text_color="#000000"]

The Rise of Surrogate Safety Data

[/et_pb_text][et_pb_text _builder_version="4.7.3" _module_preset="default"]

Collision data has long informed road safety analysis, but it is always delayed, requiring years of reported incidents.

Surrogate safety measures are non-crash measures, such as near misses that could have resulted in a crash, but fortunately did not. They are useful indicators for measuring the risk of crash occurrence by giving visibility into driver behaviour and patterns.

Because this data can be obtained faster, it enables safety experts to take a proactive approach to predicting accidents and take preventative actions more timely and efficiently.

[/et_pb_text][et_pb_image src="https://www.smatstraffic.com/wp-content/uploads/2023/04/Surrogate.Safety.Blog_.1-scaled.jpeg" title_text="Police car with emergency lights on road with caution tape" _builder_version="4.7.3" _module_preset="default"][/et_pb_image][/et_pb_column][/et_pb_row][et_pb_row _builder_version="4.7.3" _module_preset="default"][et_pb_column type="4_4" _builder_version="4.7.3" _module_preset="default"][et_pb_text admin_label="Connected car for safety studies" _builder_version="4.7.3" _module_preset="default" header_2_text_color="#0c71c3" header_3_text_color="#000000"]

Connected Car Data for Safety Studies

[/et_pb_text][et_pb_text _builder_version="4.7.3" _module_preset="default"]

Connected car data refers to data generated by vehicles that are connected to the internet, such as harsh accelerations, harsh brakes, speed, seatbelt violations and more. As a form of surrogate data, connected car data offers the benefits, such as unparalleled visibility into driver behaviours, that traditional methods can’t for safety studies and risk screening modelling. Connected car data also offers network-wide coverage without using any roadside sensors, such as cameras or lidar, making it useful for studying large networks economically and efficiently.

[/et_pb_text][/et_pb_column][/et_pb_row][et_pb_row _builder_version="4.7.3" _module_preset="default" min_height="254.5px"][et_pb_column type="4_4" _builder_version="4.7.3" _module_preset="default"][et_pb_text admin_label="Using driver behaviour for risk screening " _builder_version="4.7.3" _module_preset="default" header_2_text_color="#0c71c3" header_3_text_color="#000000"]

Using Driver Behaviours for Risk Screening and Accident Prediction

[/et_pb_text][et_pb_text admin_label="Text" _builder_version="4.7.3" _module_preset="default" min_height="166.5px"]

How accurately can driver behaviours predict accidents though? And what driver behaviours are the most important for predicting accidents? We started by studying two common driver behaviours: harsh acceleration and deceleration.

Thanks to the high-resolution quality of connected car data, we can pinpoint these two driver behaviours to any intersection or road segment, as well as isolate the behavioral data based on their relationship with another vehicle or with a Vulnerable Road User.

[/et_pb_text][et_pb_image src="https://www.smatstraffic.com/wp-content/uploads/2023/04/Driver.Behaviour.Predictions-scaled.jpeg" title_text="HELSINKI, FINLAND - NOVEMBER 04, 2016: The interior of a Tesla M" _builder_version="4.7.3" _module_preset="default" min_height="335px" height="403px"][/et_pb_image][/et_pb_column][/et_pb_row][et_pb_row _builder_version="4.7.3" _module_preset="default"][et_pb_column type="4_4" _builder_version="4.7.3" _module_preset="default"][et_pb_text admin_label="Driver Behaviours - A Study" _builder_version="4.7.3" _module_preset="default" header_2_text_color="#0c71c3" header_3_text_color="#000000"]

Driver Behaviours Matter - A Study

[/et_pb_text][et_pb_text admin_label="Text" _builder_version="4.7.3" _module_preset="default"]

To analyze whether harsh accelerations and decelerations (which we call event data) can predict accidents accurately, and their ranked importance as predictors, we did a correlation study between historic crash counts and event data using the Spearman Correlation Analysis Methodology.

Spearman Rank Correlation is a non-parametric statistical method used to measure the strength and direction of the relationship between two ranked variables. In our case, variables are surrogate and traditional safety metrics such as average speed, number of crashes, and connected car event data.

The Spearman rank correlation coefficient ranges between -1 and 1. A value of -1 indicates a perfect negative correlation, e.g., when one variable (such as average speed) increases, the other variable (such as count of crashes) decreases. A value of 1 indicates a perfect positive correlation, e.g., when one variable increases (such as average speed), the other variable also increases (such as count of crashes). A value of 0 indicates no correlation between the two variables.

We used this methodology to observe if an increase in harsh accelerations correlated with the number of crashes to determine whether harsh accelerations can be a predictor of crashes.

[/et_pb_text][/et_pb_column][/et_pb_row][et_pb_row column_structure="1_2,1_2" _builder_version="4.7.3" _module_preset="default"][et_pb_column type="1_2" _builder_version="4.7.3" _module_preset="default"][et_pb_image src="https://www.smatstraffic.com/wp-content/uploads/2023/04/Spearman1.1.png" title_text="Spearman1.1" align="center" _builder_version="4.7.3" _module_preset="default" max_width="97%" module_alignment="center" min_height="198.9px" height="121px"][/et_pb_image][/et_pb_column][et_pb_column type="1_2" _builder_version="4.7.3" _module_preset="default"][et_pb_image src="https://www.smatstraffic.com/wp-content/uploads/2023/04/Spearman.2.png" title_text="Spearman.2" align="center" _builder_version="4.7.3" _module_preset="default" width="97%" min_height="182.3px" height="100px"][/et_pb_image][/et_pb_column][/et_pb_row][et_pb_row _builder_version="4.7.3" _module_preset="default"][et_pb_column type="4_4" _builder_version="4.7.3" _module_preset="default"][et_pb_text _builder_version="4.7.3" _module_preset="default"]

SMATS looked at accident data from 2015 to 2019 from the Washington State Department of Transportation’s (WSDOT) crash data portal for 1,710 road segments in Bellevue City. Driver behaviour data (such as harsh accelerations/decelerations), sourced through Wejo’s connected car database from January and February of 2020, were also analyzed.