Sunday 22nd September 2019

Testing is critical to the safety of driverless cars

Published on June 1st, 2018

Driverless cars have been in the news on an almost daily basis this year as trials of the technology continue to take place around the world with varying levels of success. Li-Ke Huang, research and technology director, VIAVI Solutions examines the level of network testing required to safely facilitate driverless cars, and what operators must do to prepare their networks to navigate the road ahead.

Conversations about autonomous cars will continue throughout 2018 as many stakeholders battle for a place on the starting grid of this exciting and potentially lucrative technology. There are wide-ranging business cases and opportunities to monetise the technology, from more efficient logistics and fleet management to robotic taxis. Autonomous vehicles could also transform search-and-rescue missions by allowing emergency staff to enter unsafe environments without endangering a driver.

The necessary 5G engine

However, driverless cars won’t become a reality without fully functioning 5G networks. While 4G is fast enough to upload pictures, share locations or update statuses, it doesn’t afford driverless cars the human-like reflexes needed to react to real-time events and prevent accidents. 5G will be the most significant advance in data networks to date, promising ultra-reliable connectivity for a multitude of devices at low-latency. It will offer split-second response times to eradicate potentially fatal errors.

To achieve this, several network-related issues need to be addressed. The same amount of resources as are being invested in the latest flashy car technology must be spent on testing 5G networks and the protocols that will connect driverless cars. This is a substantial undertaking that will require thorough testing to ensure that recent accidents are not repeated and driverless cars are ready for large-scale roll-out.

Virtual networks; a testing reality

The scale, density and complexity of testing how a network would cope with hundreds of driverless cars in a real-life scenario, is almost impossible. Consequently, R&D efforts have moved towards virtual network testing, which can emulate up to a million driverless cars on a network to validate its performance. This includes cars moving at varying speeds and in different directions relative to network base stations.

Every single test scenario needs to be replicated, so that a virtual city of thousands of cars can be cost-effectively and safely trialled and improved. Simulators are also being developed that can support multiple simultaneous 5G air interfaces which are compatible with various wireless standards, including 3GPP’s 5G New Radio.

Li-Ke Huang

Network slicing will be a key attribute of 5G and important for R&D, representing a fundamental step on the road to commercialising driverless cars. Network slicing involves dividing a network into virtual slices to best serve the needs of specific 5G use cases. Driverless cars will be one use case and slice, although there will be sub-slices within it dedicated to different kinds of data.

For example, a car can receive safety-critical navigational information at the same time as passengers are streaming 4K videos. Network slicing will make it possible to prioritise the safety critical data over streaming videos.

Allocating network resources in this way also means that essential processes and data exchanges in an autonomous vehicle wouldn’t be affected when it drives through a built-up area or a congested stretch of road (where there may be many other devices and people using data-heavy applications). Having a network which is unaffected by the latency and capacity of others will also support vehicle-to-infrastructure communication, allowing a driverless car to respond to things like traffic lights, as well as communicating its position to the surrounding environment.

This year smart cities and smart vehicles have moved to the forefront of industry discussion. Vendors and operators may not have all the answers about how driverless cars will advance yet, but continued testing will help pave the road to the future.

The author of this blog is Li-Ke Huang, research and technology director, VIAVI Solutions

Comment on this article below or via Twitter: @IoTNow OR @jcIoTnow