Automotive Demonstrator V2: Safe Driving Indicator

posted in: Blog | 0

While road safety in the EU has improved and the number of road fatalities in the EU has rapidly decreased in the last decades, still 25.300 people have lost their lives on EU roads in 2017[1], and 1,4 million people have been injured in 2016[2]. Aggressive driving is one of the major causes of traffic accidents. Harsh driving in terms of harsh acceleration and harsh braking can immediately affect driving safety[3].

The aim of the Automotive Demonstrator V2 – Safe Driving Indicatoris to detect unsafe driving behavior such as hard braking, hard acceleration, and hard cornering, considering also environmental information, and then quantify this unsafe driving behavior within a risk score.

There are two relevant data sources for this demonstrator: vehicle driving data and historical weather data. Vehicle driving data is collected by using a data logger developed at VIRTUAL VEHICLE, based on a BeagleBone Black single plate computer connected to a vehicle’s on-board diagnostics interface. The data logger turns on at the vehicle’s start and then automatically starts recording data.

To detect unsafe driving behavior, a data-analysis pipeline is executed on the AEGIS platform. Raw data files generated by the data logger’s sensors are merged, and all trips contained in the data are extracted. In this case, a trip is the data collected by a driver between the time of engine start, vehicle operation, and engine stop. All collected trips are resampled to a fixed, regular-spaced time grid of 10Hz. Then the coordinate system of the sensors is aligned with the coordinate system of the vehicle for each of these trips.

All prepared trips are loaded to infer three common types of safe driving events, harsh acceleration, harsh braking, and harsh cornering. Driving safety-relevant events are then inferred by using the following general procedure: compute an artificial ‘event-signal’, detect safe driving events when this ‘event-signal’ exceeds a certain threshold, and then store them together with associated information on the AEGIS platform for further processing.

Trip-specific safety-relevant events (harsh acceleration, harsh braking, harsh cornering) are visualized by using markers on a geographic map. The emergence of events is furthermore captured within a driving risk score. This risk score (‘safe driving score’) is provided as a %-value in a table for all trips of a driver as well as a total score for this driver as the average value of the last 50 trip’s risk scores. Trips get exponentially decreasing weights to give more recent trips more weight.

The information generated by demonstrator version 2 is intended to be used for driver coaching. The figure below shows data from 29 trips conducted by the anonymized driver1. The table shows trip duration, number of hard acceleration events, number of hard brake events, number of hard curve events and the trip specific risk score. For instance, the first trip with the id Trip_053 has a duration of 1,13 h, 27 hard acceleration events, 32 hard brake events, and 13 curve events, leading to a trip-specific risk core of 60,95. A risk score is always between 0 (vey unsafe) and 100 (very safe). A risk score of 60,95 means this trip is safer than 61% of all trips of all drivers processed on the AEGIS platform.  Furthermore, a total risk score for this driver is generated, too.

Figure 1: Trip-specific risk scores and total risk scores of a driver

In a coaching session, this trip can be visualized to show the geo-location of the inferred safety-relevant events on a map and to discuss them with the driver. Thereby different colors indicate different types of events (red for hard braking, blue for hard accelerations, and black for hard cornering).

Figure 2: Marker- visualization of driving safety-relevant events based on vehicle data

To motivate the driver to drive even more safely, a ranking of the safest drivers by using the total risk score can be provided, too.


[1]Communication from the Commission to the European Parliament,

[2]Annual Accident Report 2018,

[3]Why monitor harsh braking and acceleration,