By Jay Nagley, Head of Business Development, Land Mobility & Northern Ireland at ANGOKA. This article was originally published in TaaS Magazine’s Issue XII 2021.

2020 was meant to be the year when autonomous vehicles were to hit the road in serious numbers. In 2017 one analyst talked of a “conservative estimate” of 10 million autonomous cars by now – about 10 million more than are in circulation a year after the deadline. With the inevitable exception of Elon Musk, every car company has dialled back the predictions of autonomous vehicles it made in 2016/2017.

Meanwhile, a lot of money has been quietly flowing into the unglamorous end of vehicle automation: Level 2 autonomy. Five years ago, Level 2 automation (combined driver assistance features like adaptive cruise control and lane keep assist) was seen as little more than a road-sign on the journey to full Level 5 autonomy. However, the cost and complexity of a vehicle having to make all the driving decisions has turned out to be far more difficult than envisaged a few years ago. 

Speaking to Spectrum magazine, Gill Pratt, CEO of the Toyota Research Institute in California, said that rapid advances in the first stage of vehicle automation (image recognition) made researchers  think that the next stage (prediction and decision making) would advance equally quickly. In fact, there is little evidence that these areas have advanced significantly in the last five years. 

A recent demonstration of UK company Five AI’s Level 4 Ford Mondeo in South London provided a good snapshot of the current state-of-the-art developments. It was hugely impressive dealing with situations it had learned to handle, taking a perfect line through one junction that would have impressed Lewis Hamilton. However, predicting what other drivers were about to do was more difficult: one approaching driver that slowed down for a roundabout made the Mondeo think it should proceed, but when the other driver accelerated slightly, the autonomous Mondeo did an emergency stop. This demonstrated the inevitable overreaction of a system that is rightly programmed not to have full confidence in its own decisions. 

Instead of chasing the rainbow of Level 5 autonomy, the global Tier Ones like Bosch and ZF (the companies who actually develop most electronic technology in the car industry) are now concentrating on two areas for on-road automation: Level 2 Plus for the entire industry, and Level 3 for premium models. 

Level 2 Plus is a relatively recent concept, which is about stretching Level 2 systems as far as they can go, while still leaving ultimate control with the driver. German giant ZF has probably done most to popularise the concept, which tries to provide assistance for almost every manoeuvre a driver makes, even including “merge assist” for motorway slip-roads. The company says it can provide a complete system to an OEM for well under $1000 – a fraction of what an autonomous system would cost.

Level 3, which was originally framed to mean conditional automation, where the car could drive itself, but would hand back to the driver if it encountered a situation it could not cope with, has been subtly redefined. Today, it effectively means automated driving on a highway at relatively low speeds (up to 50 km/h). The reason for restricting it to highways is that a Level 3 system would need to stay in control for the 5-10 seconds a driver needs to resume command – so it can only operate where the possible challenges during that time are limited.

So, does the march of driver assist systems mean the marginalisation of autonomy? Not necessarily, but it means full autonomy will only to be used in specific use cases for the next 5-10 years. The first application is likely to be logistics – so-called ZEAL (Zero Emissions Automated Logistics). Vehicles that operate in bounded environments (e.g. ports or logistics hubs) are far easier to automate, as their entire operational domain can be mapped, and interactions with other vehicles can be controlled. 

Currently, most of these autonomous vehicles are actually teleoperated, with a controller overseeing their activities. This hugely reduces cost and complexity, as the “brain of last resort” is a human one, which has had millions of years of free training in handling unexpected situations, as opposed to a hugely expensive computer whose evolution has to be paid for. 

However, this does raise another issue. Automated logistics vehicles need to be in constant contact with both their operator and with each other. The communications themselves are well understood, and constantly improving (5G being the most important advance), but the greater volume of communications creates major opportunities for hacking. Quite often the cybersecurity aspects of trials or deployments are left until the last minute – the boring stuff that nobody wants to take charge of. It is also expensive, as operators must not only set up the security system but also to keep it regularly updated with fresh cryptographic keys and Certificates to refresh the system and communications security. 

This is the issue that Angoka has addressed with its hardware-based cybersecurity solution, using Device Authentication Units (DAUs) to handle the identity generation and management of the device, as well as its authentication and authorisation. As it is based on hardware, it is intrinsically safer than relying on software alone. It is also more cost-effective as the DAUs collaboratively create passwords in real-time, removing the need to periodically load new passwords to maintain security. Finally, it is widely applicable, because it provides trusted communications in an untrusted network.

When the vehicle is started, each device contributes to the creation of a new session key – a unique password shared between authorised devices on the network. Effectively it is like providing a police escort to a prison van to ensure that the van arrives securely at its destination (unless it is an episode of Line of Duty, of course).  

This solution is already being used in ZEAL trials. It won’t magically solve the problems of delivering Level 5 autonomous vehicles in the 2030s, but it does simplify the issues around implementing Level 4 logistics solutions in the 2020s.