Data Highway for the Car

IAV examines the use of Ethernet in cars as part of a BMBF project

New sensors such as video cameras or 3D lasers generate quickly increasing volumes of data with high real-time requirements that cannot be transmitted via the CAN bus. Automotive Ethernet is a viable addition or possible successor to the CAN bus. The standard permits large data transfer rates, has proven its worth in the IT world and could be the basis for interconnection in the vehicle in future. A test vehicle shows what the new technology can do.

The CAN bus reaches its speed limit at the latest with one megabit per second (Mbps) per segment. Today already that is no longer sufficient to connect a large number of control units on one single bus, which is why typical cars such as the VW Golf or VW Passat have half a dozen separate CAN buses, each responsible for different vehicle domains, such as the engine, chassis, infotainment or driver assistance systems. Most data communication runs within one domain, the rest is handled by gateways to bridge between the domains.

Although additional bus systems such as MOST and FlexRay have recently found their way into the vehicle, their purpose is by definition not as general as the CAN bus. FlexRay, for example, is used for chassis or engine control, while MOST is specialized for handling infotainment systems. To bypass the limits of current bus systems, many car manufacturers therefore use proprietary solutions for connecting new sensors with the vehicle. “In some cases, camera-based driver assistance components use LVDS (low-voltage differential signaling) or analog connections to the driver assistance systems, or Ethernet is used as a proprietary point-to-point connection manager”, explains René Röllig, senior project leader for driver assistance systems at IAV. “But it would be better to have a uniform standard also for high data transmission rates.”

Strong real-time requirements in the vehicle

It is already clear which requirements a new standard will have to meet. It has to allow high data rates well in excess of one Mbps together with high transmission reliability and low jitter (unwanted delays in messages compared to the system clock). It is important that the system can fulfill the strong real-time demands in the vehicle. “Safety-relevant signals to the brakes or to the chassis have to be transmitted within strictly defined time tolerance limits”, says Röllig. “The CAN bus has always been able to cope with these requirements in its speed class, giving priority to the most important message. A successor will have to be able to do the same.”

One promising candidate is Automotive Ethernet, an offspring of the well-known IT transmission standard. “This new technology is just emerging”, reports Röllig. “However, at the moment only a few manufacturers are producing the right sort of chips.” While in IT computer networks data rates of above 10 gigabit per second (Gbps) are state of the art, the current automotive Ethernet version manages only a maximum of 100 Mbps. This is because of the simple, light-weight and cheap twisted pair cables used for data transmission, compared to the IT sector which mostly uses shielded cables with four or eight conductors or fiber optic cables. On the other hand, the great advantage of the Automotive Ethernet is that it allows to use the same cables as CAN. Above the physical transmission layer, data communication is handled by the well proven internet protocols TCP/IP (transmission control protocol/internet protocol).

Time-triggered Ethernet allows guaranteed delays

However, the communication protocols have to be adjusted and enhanced in advance to the real-time requirements in the vehicle. Here, methods like reduced ethernet frame sizes and a time multiplex procedure to control access of the individual control units can be used. Various static rules for TDMA (time division multiple access) let the TTE (time-triggered Ethernet) guarantee reliable data transfer and predictable delay times. “Another key topic is the network architecture”, says Röllig. “A conceivable solution consists in a star topology with multi-port switches distributing the data. Alternatively, every control unit could be equipped with two ports as input and output to connect the units in a chain.”

IAV is currently working together with Hamburg University of Applied Sciences (HAW) to examine what the data backbones in future cars could look like. In the framework of the RECBAR project (real-time Ethernet in car backbones) funded by the BMBF (German Federal Ministry of Education and Research), the partners have fitted a Golf 7 with additional sensors such as a HD camera and 3D laser scanners which are connected to the vehicle by TTE. “Their high data transfer rates generate great stress in the system”, says Röllig. “In this way, we can see whether the Automotive Ethernet is capable of coping with these requirements.” Gateways are used to exchange messages between TTE and the CAN buses so that the researchers can also examine whether they reach their destination with sufficient priority.

Simulations are used to model critical system states

To examine complex network architectures, IAV and HAW also develop simulation methods to model the TTE components and permit detailed analysis of communication. “The results obtained up to now show that data transfer fulfils the basic requirements”, reports Röllig. “In future, we will deliberately provoke critical system states and simulate various conceptual variants. So it is for example possible to route data via an additional redundant link and to model failure scenarios by switching off certain services and subscribers.”

The project is continuing until mid 2016. Röllig doesn't expect to see the first standard use of the Automotive Ethernet beyond proprietary niche solutions before 2018.