Backend data aggregation complements sensors in automated driving
The basic assumption for the Ko-HAF research project is that at high speeds in relatively constant environments like long-distance highways, the driver does not need to monitor constantly the car; he can dedicate himself to other activities, however he must be able to take over the wheel in few seconds if necessary. To enable such a scenario, it is necessary that the vehicle’s perceptional horizon goes beyond what is possible with local sensors. Here is the focus topic of Ko-HAF: The project aims at establishing a virtual data environment in which the vehicles can move safely and that enables them to early identify possible obstacles.
Towards this end, the vehicles transmit data about their environment to a backend server. There, they are collected and compacted so that the vehicles have a current, accurate map available, which provides the required foresight in the sense of an artificial horizon.
During the first half of the project, the focus was on the design and development of the Safety Server as well as the two-way communication between vehicles and Safety Server. Upon request, the vehicles receive an accurate digital map and can upload their environmental perceptions – static traffic objects such as signs, markings and cetera. Aggregation procedures to fuse the numerous messages from the vehicles in the server are currently being developed.
On the vehicle side, self-localization and georeferencing of objects have been implemented. The participants also jointly developed a method to integrate the data from different vehicles into the server-side environment model. This was a non-trivial undertaking, the Ko-HAF partners summarize, because the participating vehicles were equipped with different sets of sensors which provided data in different formats. The tests were carried out not only in the laboratory but also “in the wild”, on the basis of real data.
Another topical focus was human-machine interaction during autonomous driving. In the case of highly automated driving, it is anticipated that the electronic chauffeur takes control of the car and the driver can turn to other tasks. This makes it necessary to be able to assess the driver’s availability and to know how quickly a driver can re-assume control if necessary. This depends on the activity the driver is engaged in; also fatigue is a major factor. In studies in the simulator and real vehicle data have been collected, which should allow modeling of the driver’s availability during highly automated trips – findings that can be used in the design of future vehicles.
In this context, the 16 Ko-HAF project partners laid the conceptual foundations for future highly automated driving functions such as automatically merging into the highway and exiting it. Likewise, detecting and avoiding hazard points, the transition to a risk-related situation in the event of system failures have been investigated in detail. The same applies to the development of corresponding test tools. For the latter main areas of work, the results are planned in the second phase of the project.
In Ko-HAF, 16 companies and research institutions are collaborating, including carmakers Audi, BMW, Daimler, and Opel as well as tier one suppliers Bosch, Continental and Visteon and academic institutions like the Technical universities of Braunschweig, Munich and Würzburg. The project is expected to be completed in November, 2018.
Related articles:
Cars learn to deal with each other automatically
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
