Driverless tech can’t rely just on connectivity

There is an argument that, to a debateable extent, autonomous vehicles require connectivity. Dr Graeme Smith, CEO of Oxbotica, thinks that self-driving cars must have the capability to operate when there is no connectivity at all.

“The design of our vehicles doesn’t require connectivity or GPS because we design the systems to be self-sufficient, so that they are robust and reliable under all conditions,” he says. The problem with having autonomous cars that rely upon connectivity alone falls down to the fact that if it fails there is no way of recovering from it – such as when a self-driving vehicle is within a weak or dead mobile signal area. “So we build everything we need to run autonomously into the vehicle itself, to make sure that it’s self-sufficient,” he adds before explaining that “the question is not whether the vehicle should be connected but what data they would have if they were connected”.

What interests him most is the situation where you have fleets of vehicles in cities and the question of how they communicate with each other. In this case he comments: “Yes they will need to be connected to each other but you can’t depend on connectivity for the operation of each individual vehicle.” Why? Well in his opinion it’s like your car today but the added bonus of connectivity is that you will be able to do more things – such as streaming music over the air. Connectivity will also allow your car to connect to the infrastructure around it – to the traffic lights for example. However, music streaming and even traffic lights aren’t essential to the operation of the vehicle.

Rickard Widerberg, chief marketing officer at Telenor Connexion, interjects to comment: “For me this is not debateable at all because fully automated vehicles need to communicate with the world surrounding them.” He says they also must be able to ensure that connectivity is essential for intelligent, improved driving and traffic management. Cars of today are already connected to support the driver with best decisions, e.g. traffic, street conditions and this will for sure also be used for automated driving,” he adds before saying that passengers demand solutions and applications that require connectivity to the vehicle. Music streaming is but one of these.

Cloud-based connectivity

Marcus Rothoff, autonomous driving programme director at Volvo, highlights that: “Connectivity by cloud-based solutions offers new possibilities for all cars to share and get access to dynamic data beyond the horizon and for autonomous driving this is a must: one example on probe-sourced data is the ‘Slippery Road Alert’ and ‘Hazard Light Alert’ functions we launched last year.”

Ian Riches, director of Strategy Analytics’ global automotive practice, nevertheless, supports Smith’s view as he argues: “There is no immediate need for connectivity to enable automated driving at low speeds and this has been proven by a number of driverless shuttles as well as by Alphabet, Waymo and Tesla.” He also finds that connectivity and autonomy have been to date on separate development paths. “This is about to change with the onset of direct device-to-device communications, which will introduce the mesh networking effect and V2X propositions” that will extend the range of on-board sensors. The mesh networks will involve other vehicles and connectivity to infrastructure.

Smith explains: “It might be useful to connect driverless vehicles to insurance services and to manage them as a fleet. In our implementations we’d probably do V2V, which means connectivity to anything else such as infrastructure. When the car is driverless you want to control your vehicle centrally and to share hazard or mapping information between vehicles.” While the data from autonomous and semi-autonomous vehicles is expected to increase over time, he claims that he wouldn’t expect this to see 2 petabytes of data coming off these kinds of vehicle. If this were the case, then he’d suggest that there may be a design fault because it shouldn’t happen at all. The vehicles will have their own limitations and, to a certain extent, this can’t be avoided.

Top priority: security

Widerberg remarks: “We are in an environment where lives are involved and so security can’t jeopardise safety. That does not necessarily mean that functionality is removed or reduced, but for sure this means we need to design the services and communication in a smart and in a secure way from the start. We also need to be able to securely update firmware and other services to avoid security issues.” Subsequently he thinks the communication and the services constantly need to be monitored for abnormal behaviour, indicators for malfunctioning, security attacks, etc.

“Telenor Connection has developed our data analytics service called ARTS: We are helping our customers to detect patterns, abnormalities and potential misuse in the connectivity early based on machine learning,” he explains before adding that his firm aims to proactively help its customers by acting upon the data insights it gains from the service.

Therefore, the way to get ahead of the game is not to necessarily have an always-on autonomous vehicle that is always connected but to have a means of allowing the vehicle to be self-reliant while allowing data and applications to be updated when a connection exists. With this approach the car when it is beyond the reach of the IoT will still be able to operate, drive or be driven. Beyond this what is most needed is some common standards about security, privacy and responsibility. Smith concludes that there is no need for a balance between security and functionality because if it is managed correctly there should be no limitations. So IoT’s importance will still grow.

[Auto.Jarvis.2017.02.09]