VTT Technical Research Centre of Finland and mobile giant Nokia are leading a consortium of Finnish companies and research institutes that have joined forces to figure out how 5G networks can transmit information to and from vehicles while on the road. To do this, they are testing the capabilities of 5G network technologies when combined with VTT’s robot car, called Martti.
The researchers, as part of the 5G-SAFE project, are studying what kind of novel road safety services, such as monitoring road conditions and gathering real-time incident information, that 5G will enable for supporting autonomous driving.
These road safety services will rely on sensors and Internet of Things (IoT) devices to collect data from a self-driving car’s lidar, radar, video systems, as well as information from roadside infrastructure such as weather stations, traffic cameras, and traffic lights. These devices will communicate over 5G networks to cloud-based services that use algorithms to process information and subsequently deliver alerts to other vehicles in the area.
Researchers have been working to bring the data processing closer to the vehicles by the means of edge computing for reduced latencies and improved scalability, says Tiia Ojanperä, senior scientist and project manager at VTT. Now they are implementing the solutions into real 5G test networks and vehicular platforms for validation.
“We were running the first pilot in June, where VTT’s robot car ‘Martti’ was connected to a 5G test network available in a vehicle test track in Sodankylä, Finland,” said Ojanperä said in an email interview. “The test network in question was still pre-5G, and waiting for real 5G capabilities that will be deployed once the equipment (network devices and user terminals) is commercially available.”
In the pilot, Martti was used to test the ability to detect obstacles and grooves in the road by the means of collaborative sensing. Another test vehicle transmitted 16-layer lidar sensor data on a 12.5-hertz frequency to VTT’s MEC server located in the 5G test network. The MEC server hosted an algorithm, which optimized Martti’s route.
The second and most recent test VTT and Nokia jointly conducted, in August, involved communication of Martti’s sensor data to a server via Nokia’s prototype 5G equipment.
For as long as 5G networks have been on the horizon, observers have speculated about the main applications for these new networks, including certain automotive applications. Recently, experts have questioned whether it’s practical to use 5G for autonomous vehicles when mobile coverage of remote and rural roads is so limited.
The 5G-SAFE research project has taken this coverage limitation into account as part of their strategy in building out the technology.
“In the project, we assume that multiple radio technologies (5G, 4G, ITS G5, or even satellite) will eventually be used in implementing the services, as not all the 5G capabilities (low latency, high bandwidth, etc.) will be available everywhere the vehicles travel,” says Ojanperä. “The idea is to select the most appropriate means of communication dynamically among the available ones. Also, the services will need to be adapted to the network capabilities, and in some remote or rural areas, only a subset of the features and information may be available.”
While this strategy may address coverage issues, questions remain about what aspects of driving could be safely handled over a 5G network as opposed to within the car's computers. Ojanperä believes that 5G-enabled distributed cloud services, such as the ones developed in the 5G-SAFE project, and direct vehicle-to-vehicle communication, can provide autonomous vehicles sufficient knowledge of their surroundings—enough for them to operate safety.
“One of the biggest benefits [of 5G networks] is edge computing and its enabled distributed crowdsourcing opportunities,” said Ojanperä. “Edge computing is not access technology dependent, but the broad bandwidth of 5G makes it useful by enabling real-time local dynamic map updates and communicating even raw sensor data between vehicles via intelligent infrastructure.”
Ojanperä concedes that vehicles will still need to function without this additional information if the connection to the services is lost. She says that in those situations, the vehicles will have to rely on their own sensors with a limited range or implement some fail-safe functionality, like stop operation, when connectivity has been lost.
With the test environments and communication solutions in place, the next step will be to develop more advanced intelligence on top of the collected vehicular data, according to Ojanperä.
Ojanperä added: “It is not a big secret that real-time and reliable data is the oil which enables machine-learning algorithms to make cars and transport systems ever more intelligent. Connectivity is the key enabling technology for pushing forward the transport automation megatrend.”