June 17, 2016: Video from AMIDST
On April 6th and June 17th, 2016 several project members of Daimler AG and RTD Manager and Deputy Coordinator Anders L Madsen (HUGIN EXPERT A/S) took a Mercedes Benz experimental vehicle for a drive on German highways to evaluate developments made in the AMIDST project in real live traffic situations.
As one of the use-case providers in the AMIDST project, Daimler AG considers the challenge of analysing real and highly complex streaming traffic data using the algorithms and tools developed in AMIDST. The aim of the use-case is to recognize traffic maneuvers of the own vehicle and its surrounding vehicles in complex highway traffic scenes.
Recognition of maneuvers in live highway traffic involves analysing situations that develop quickly over time using information captured in the order of milliseconds by several sensors – often featuring several vehicles. A test drive of, for instance, one hour requires that several millions of database records are analysed online by the AMIDST models and algorithms. These models and algorithms must run under severe constraints on the time and memory used as required by the target platform. Hence, AMIDST also considers the challenge of satisfying the requirements on memory, speed and computation power induced by the computer in the car.
The purpose of the test drive was to evaluate the performance of new and dynamic models developed as part of AMIDST as well as to collect additional data from new sensors. The models showed promising results with improved performance compared to previous models.
The work in AMIDST will continue to improve the performance of the model and the supporting algorithms for analysing large amounts of streaming data.
Please follow this link to see the video of the drive or copy the link address to your browser to download the mp4 video file. Driver Galia Weidl, co-driver Anders L. Madsen.
Explanatory text for the video: The video demonstrates an early recognition of maneuvers in a real highway traffic as evaluated by algorithms, installed on-board of the own experimental vehicle. The test drive is shown from two perspectives:
- A view on the highway-situation from inside the own vehicle;
- Visualization (in the system-display) of maneuver recognition performance (computed by the own vehicle) for both vehicles, the own and for one in-front.
The early recognized intention of maneuver is visualized by a blue arrow, which here represents the cut-in maneuver-intention of an in-front vehicle. Its actual start of the Lane Change Maneuver, i.e. approaching of the lane marking, is visualized by a red arrow.
The AMIDST toolbox is available at GitHub: http://amidst.github.io/toolbox/.