Emmanuel Bert describes how our Real Data Set Consistency Checker can help you in calibration and validation tasks.
Why
The ability to create and import a Real Data Set (RDS) in Aimsun Next is very useful for calibration and validation tasks. Given that the objective of any model is to be as close as possible to real life, the RDS checker allows you to import real traffic data to the model to make direct comparisons and adjust internal parameters to fit these values.
Real data sets are essential elements of an Aimsun Next model, but sometimes the quality of the data provided by the client or other data sources is less than perfect and may have gaps and inconsistencies that are often difficult to detect, especially in a large-scale, complex network with a high number of detectors.
Additionally, when using several sources of data, you are quite likely to observe variations, so you need to be sure that the different data sources are compatible in terms of data type and date/time.
Of course, trying to calibrate a model based on a real data set with inconsistent traffic values will make the process more difficult, error prone and even impossible – for instance, you cannot satisfy two inconsistent traffic values in the same section.
What
The Real Data Consistency Checker was introduced as a Fast Track feature in Aimsun Next version 8.4.1. This tool can apply several consistency checks based on flow, speed and occupancy (if available) to either detectors or stations.
Individual consistency:
- Data checking: detection of missing, negative, ‘nan’, erroneous data (for instance Occupancy > 100%).
- Min Excess of Measured Speed vs Speed Limit: consistency of the speed values.
- Min Excess of Measured Flow vs Capacity: consistency between the RDS flows and the section capacities defined in the model.
Congestion consistency:
A congestion consistency message will appear when these two criteria are fulfilled simultaneously:
- Min Occupancy to Identify Congestion: occupancy threshold to start detecting the congestion.
- Max Flow to Identify Congestion: flow threshold to start detecting the congestion.
Spatial consistency (count or flow):
The algorithm checks the difference between two measure points that have no interference in between (no merge/diverge nor centroid connections) or when the incoming vs outcoming flow differences in the node and applies the following criteria:
- Min Flow Value to Check Consistency: minimum flow to consider, to avoid unwanted detection with very small flow values.
- Max Distance to Check Consistency: distant between objects to consider.
- Min Flow Difference to Report Inconsistency: acceptable difference between diverging values.
Things to note:
- Find the possible difference of flow due to distance between the measuring points by calculating the approximate storage capacity.
- When the record corresponds to a partial value (not covering all the lanes of the section), data on the missing lanes will be sought within a range of 50m. If nothing is found, this record will not be checked for spatial consistency.
You’ll find more detail about these parameters in the User Manual: qthelp://aimsun.com.aimsun.8.4/doc/UsersManual/RealDataSetEditing.html#checking_real_data_consistency.
How
After defining the real data set in the Project menu, select Retrieve & Check:
In the Consistency Parameters tab, modify the criteria for data consistency checks:
Click on the Check button to generate the list of inconsistencies in the Results tab.
To check with different time aggregations, just restore the RDS with a different time aggregation before performing the check.
Output
In the Results tab, you will see the list of consistency messages based on the RDS retrieved:
Example of messages regarding individual consistency:
They contain the line number, warning sign, the type of check (consistency type), the detector, the time of the record and the message explaining the issue detected.
Example of messages about spatial consistency:
The following message shows an inconsistency between 2 detectors that have no interference in between:
And finally, consistency issues for a node: