MobiArch: End of congestion warning

The project develops a procedure for the determination of traffic jam ends and the warning of road users. Road users should be warned for half a minute and at a distance of up to 1 km before reaching the end of the traffic jam in order to avoid rear-end collisions. The focus is on temporarily jam-prone sections on multi-lane highways, especially on construction sites. A process model is developed that covers all steps from data collection to issuing a warning. The ends of traffic jams are to be determined as well as their positions are to be predicted with the help of propagation models for jamming ends. Data sources include temporarily installed sensors, in particular Bluetooth and radar sensors, floating car data and existing cameras. For the output of the warnings, the integration in third-party applications (apps) is specified. To test the processes, it is planned to use data collected in the course of a test field at the construction site Enztalquerung on the A8 near Pforzheim.

Bluetooth and radar sensors and Floating Car Data (FCD) provide different traffic parameters such as speeds, volume, travel times. Each of these traffic data sources has its inherent strengths and weaknesses and may not accurately represent the actual state of traffic necessary for modelling and predicting traffic conditions. In this project, the data sources are used to complement each other through data fusion to provide accurate data for traffic congestion detection and end of congestion tracking.

The first part of the project involves real-time data acquisition from radar and Bluetooth sensors placed on the highway and FCD data from INRIX, HERE and TOMTOM. Spot speeds, traffic volume and road occupancy from radar sensors will then be fused with travel times and mean speeds from FCD and Bluetooth datasets. The result of this fusion process will be traffic flow rates and vehicle density characterizing the traffic conditions. The fused results will be the foundation of the end of traffic congestion detection and tracking stage.

The core results of the project include a process model that specifies all steps from data collection and data fusion to the calculation and output of an end-of-congestion warning, as well as algorithms for the detection and tracking of the end of congestion, which can be used to warn road users when they are approaching the end of a congestion on multi-lane motorways or federal highways. The detection of the end of a traffic jam is carried out with a high spatial-temporal resolution of 250 m and 15 seconds, which allows road users to be notified on time and in real time. A specification describes the technical requirements for third-party applications to process the data and output it as a message (traffic jam warning app). This is the cornerstone for existing apps to integrate warnings of the end of traffic jams as a functionality in the future.

Project results in detail (in German): https://mobidata-bw.de/img/Toolbox_Img/04-MobiArch-Steckbrief-Stauende-Alarm-Final.pdf

The testing and validation of the developed tools will take place in 2021 and 2022 in the course of the six-lane expansion of the A8 between Pforzheim-Nord and Pforzheim-Süd. For the acquisition of (local) traffic measurements related to the road cross-section, radar detectors with lane-separated acquisition will be set up as part of the integrated congestion warning system (iStWA) A8. These should detect data at least in the 15-s interval according to TLS 2012, type 52/116 (sample up to 99%). The results serve as a basis for direct congestion detection in the sub-segment and for the congestion tracking model. In addition to the local detectors, there are also route-related combination detectors (Bluetooth BT, Bluetooth Low Energy BLE and WiFi technologies), which are mounted on set-up devices of the congestion warning system in the test field and in particular detect travel times and congestion for the sections (sample 30 to 70%). As a third component for determining congestion, the route-related traffic flow and congestion data from the country's inventory (FCD data source Inrix, among others, where available) fused to 250m sub-segments are used. WebCams and other map-based traffic information are used for validation. AI video detectors are added as a reference to include their potential for detecting the end of congestion.