Cooperative Highly Automated Driving
The research project Ko-HAF – Cooperative Highly Automated Driving – aims at the next major step towards self-driving cars: highly automated driving. A highly automated vehicle has the ability to control movement laterally and longitudinally with no permanent driver supervision. However, the driver must always be in a position to resume control within some advance warning time after being alerted.
First generation highly automated driving is intended for low to very low speeds in simple traffic situations such as congestions on motorways. It controls the vehicle autonomously based on the car's built-in sensors.
The second generation, which is the focus of Ko-HAF, will be able to deal with higher speeds on well-built traffic infrastructure, in order to improve safety, efficiency and convenience in this common use case.
A core component of the project is a layered road map, which includes both a high-accuracy model of the road environment and short-term information about current events (e.g. information displayed on variable message signs, accidents or congestions). This map represents features such as lane markings, painted symbols and traffic signs with centimetre accuracy. A "safety server" is responsible for storing and managing the map. Software for this server is being developed as part of Ko-HAF, with FORWISS as the main contributor. In the conceptual framework of collective perception, the map represents the shared memory of the vehicle fleet, allowing each vehicle to anticipate circumstances beyond its own sensor range.
In order to ensure reliable performance, the map has to be updated continuously. To achieve this, Ko-HAF is enlisting the help of the vehicle fleet. If the sensors of a vehicle detect differences between the map and the perceived reality, that information is communicated to the server via mobile radio networks. The server then combines this data with observations from other vehicles to produce an updated version of the map, which is shared back to the participating vehicles – again using mobile radio. The development of algorithms for conflating observations from a diverse fleet of vehicles offers a wide range of research avenues. FORWISS will contribute algorithms specialized on longitudinal features, such as lane markings.
More information is available from the official project website.