EVW in Barcelona

Barcelona firefighters have different stations in the city to optimise their actuations. Still, Barcelona is a busy area and traffic can get challenging for fire trucks in emergency. The Emergency Vehicle Warning service informs road users about emergency vehicles from the fire brigade approaching, even before seeing and hearing them, so that they can anticipate and give way in a safer and more efficient manner.

Sample Data Description

For the evaluation of the EVW service deployment in Barcelona, the relevant log items are the denmaction, denmevent and CAM log files. All of these log files are automatically recorded by the PID of the user. The denmevent logs contain the basic information for each event (e.g. timestamp, log_stationid, eventcausecode etc.). The denmaction logs contain the status of the HMI and thus record the information that was presented to the driver at each moment in time with a frequency of approximately 1Hz. The following fields are particularly relevant for the RWW service: 

• Warningdistance: indicating the distance to the event location 
• Servicecategorycode: indicating the message displayed on the HMI 
• Pictogramcategorycode: indicating the pictogram displayed on the HMI 

The eventcausecode field from the denmevent tables, in combination with the servicecategorycode and pictogramcategorycode fields from the denmaction tables, enables the evaluator to differentiate between the different services using the same message types (DENM). The following combinations of values defines the EVW service deployed in Barcelona: 

The CAM logging is mainly used to derive the vehicle position, speed and acceleration for the entire duration of the event, also at a frequency of approximately 1Hz. For a more in-depth description of the logging format and the data fields in each log type please visit Open Common Log Data Format.

The example data set provided here consists of all EVW events logged in Barcelona. The data set includes both baseline and treatment events. ‘Baseline’ indicates that the service was not active (no information was presented to the driver) and is used as a point of reference, whereas ‘treatment’ events are logged when the service is fully functional. Separating events by baseline/treatment is possible by analyzing the content of the denmaction log files and, more specifically, by checking for the existence of rows where eventmodelid==6 (indicating that certain information was displayed on the screen). The App by IDIADA deployed in Barcelona randomly switches between baseline/treatment with pre-defined probabilities (percentage of baseline events is 25%). Therefore, the example data set provided here consists of roughly 150 treatment events and 50 baseline events. 

To simplify the provision and processing of data, all denmaction logs are grouped into one iviaction.csv file, all denmevent logs into one ivievent.csv file and all CAM logs into one CAM.csv file. Individual events can be identified through unique combinations of the (log_stationid, eventid) fields. As defined in the log format specification, each event should begin with eventmodelid=5 and eventactionid=1 and end with eventmodelid=5 and eventactionid=3. In case there are multiple instances of these combination for one event, the first (5,1) and (5,3) are used to define the start and end of the event. 

You can download the log data here.

Evaluation Results

Several indicators can be used to evaluate the impact of EVW on the traffic flow, mainly in terms of safety.
The following table briefly describes some of the main indicators. EVW is expected to increase the maximum deceleration (since these are negative values it is a decrease in absolute terms) and the frequency of hard braking, since drivers are informed in advance of the road works and should brake (or at least coast) from further out. Furthermore, the speed variance should also decrease, both within each event and between all vehicle passages.

The evaluation results data set provided here also contains a table with the above-mentioned indicators calculated for each event. The overall impact of EVW can then be calculated by grouping events by baseline/treatment and comparing the indicator averages over both groups. The following table gives an overview of the main indicators as extracted from the sample data. The results based on the sample data indicate that, when using the service, there is more hard braking and the maximum deceleration decreases. At the same time, the distance of the first braking from the event increases and the mean speed of the vehicles at 100m distance to the event decreases, which has positive implications on the traffic safety.

To better understand and evaluate these indicators, the mean values are not enough. The probability density functions (PDF) of the indicators for baseline and treatment should also be considered. The following figure illustrates this concept for two of the indicators. Although there is a clear difference in the mean values between baseline and treatment, the relatively widely dispersed distributions and high standard deviation values suggest that the impact of the service is still uncertain.

In order to have more robust and meaningful evaluation results, a larger data sample and a more in-depth analysis of the data (e.g. filtering events by certain criteria, taking into consideration peak/off-peak traffic conditions, checking for incomplete events or errors in the log data etc.) is required. For further details on this please consult <add link to D6.4 (when it will be published)>.

You can download the evaluation data here.