Safe Automated Driving: Collaborative Project Starts

Automated and connected driving is one of the automotive sector’s technologies of the future. For fully automated vehicles to be underway on our roads, guaranteeing that they are steered safely in every situation is essential. This becomes especially difficult when the vehicle hands control back over to the person at the wheel. This “takeover” time, which the driver needs to get his or her bearings in a critical or complex situation, is usually calculated at ten seconds. In many traffic situations, that is far too long: At a speed of 50 km/h, a car travels almost a further 140 meters in city traffic in this time.

This is the starting point for the new project: By predicting how the traffic situation is going to develop, the vehicle itself should be able to anticipate and plan its maneuvers. If the information from its surroundings is evaluated and interpreted properly, it is possible to align a vehicle’s individual motion planning with that of other road users. However, current approaches to automated driving, which work with conventional rule-based algorithms, are unable to reflect adequately how the traffic situation is going to develop. In a finite set of rules, the possible interactions of all road users as well as consideration of the road infrastructure and all contextual information can only be expressed incompletely.

Modern technologies help with representation of the environment

That is why the project partners are first of all working to develop a representation of the environment that contains all important aspects of the traffic scene: All static and dynamic objects and the infrastructure in the vehicle’s range of perception. Already today, modern sensor technologies are delivering more and more information that can be used for this purpose: Radar, cameras, and sensors for optical measurement of distance and speed. To reflect that the traffic situation changes over time, observations are accumulated over a certain period. In addition, the representation should also already produce a basic interpretation of the situation, e.g. take speed limits into account.

The description of the environment is the starting point for the prediction: On the basis of the accumulated measurements from the observation period, artificial intelligence is used to train a model that looks into the future. The possible courses of future situations are then used for the decision algorithms needed to support and improve the vehicle’s maneuver planning. A test vehicle will be developed and used within the project to evaluate the predictions under real conditions.

About the KISSaF project:

“KISSaF – AI-based Situation Interpretation for Automated Driving” will run from 1 January 2021 to 30 June 2023. The total project volume is about € 4 million, of which 68 percent is being funded by the Federal Ministry for Economic Affairs and Energy. Collaborating in the project are three companies, ZF Automotive GmbH Germany in Düsseldorf (project coordinator), ZF AI Lab in Saarbrücken, and INGgreen GmbH in Koblenz, as well as the Robotics Research Institute of TU Dort­mund University.

Back