Autonomous vehicles: myth or reality?
Head of Strategy & Portfolio, Atos
Vice President & Global Market Leader: Transportation & Hospitality, Atos UK&I
Posted on: 22 October 2019
The vision of our towns and cities full of driverless cars has captured the public imagination. Autonomous vehicles have been hailed as the answer to our traffic problems, with accidents and congestion eliminated on calmer and quieter roads and people freed up to watch TV, work or even sleep on the move. But is this vision of the future realistic and when is it likely to happen?
The likelihood is that we will see an evolving hybrid mobility mix of Connected and Autonomous Vehicles (CAVs), comprising both autonomous vehicles and connected vehicles driven by people. A useful reference point is the set of globally recognized levels of autonomous driving systems set out by the Society of Automotive Engineers. Level 0 vehicles have no autonomy and must be controlled by a human at all times. Level 5 vehicles are fully autonomous and don’t require any help from a human to operate.
Society of automotive engineers (SAE) automation levels
Levels 1, 2 and 3 (with continuous supervision by drivers) are already available or will be available in the next few years by almost all vehicle manufacturers. They have major benefits for public safety (and insurance premiums) through more controlled, informed and safer driving. Level 4 driving systems (autonomy within certain parameters) have been around for some time in isolation: Rotterdam Docks, for example, uses completely automated vehicles within a geofenced site. However, the realities of introducing autonomous driving at scale and within a hybrid traffic environment bring a number of key challenges.
Computing at the edge
Autonomous vehicles require huge amounts of compute power: the technologies cannot make critical decisions in real time if they rely on cloud-based processing. There is simply too much latency that may cause unacceptable and dangerous delay.
Developing sufficient computing power is a work in progress, with edge computing at the heart of the solution. Edge computing helps transfer computing power out to the edge of the network, where devices collect and process all kinds of direct and indirect data and visual information required for effective and safe autonomous driving (including object proximity, traffic signals, pot holes and weather conditions).
When these devices are mobile, they can form an intelligent swarm. Just as, in nature, swarms of insects or murmurations of starlings coordinate their interactions, so too we can expect the Internet of Things (IoT) to enable multitudes of entities and devices to combine, forming dynamic and intelligent collectives. In this way, a swarm of driverless cars could interact to move safely and efficiently at speeds that are just not possible if local control decisions are made in isolation. Vehicles could join or leave the swarm as needed, contributing their own localized insight to traffic and environmental conditions.
Smartphones already act as rudimentary swarms: they crowd-source traffic information by detecting when phones are bunching up and slowing down – but the vehicle driver still needs to decide how to act on such insights. Autonomous vehicles interacting collaboratively will generate a much broader view.
Security and safety
As connected vehicles enter the mainstream, so too does the potential for their malicious exploitation. There could be significant impacts on personal and public safety if hackers gain remote control of a vehicle or adversely impact its behavior.
Any IoT-enabled system tends to lead to an extended ‘attack surface’, so autonomous driving systems require more dynamic, scalable, decentralized and intelligent security mechanisms than less interconnected operating systems. Technologies like blockchain could be a means of immutably identifying autonomous objects, validating the authenticity of individual pieces of software and establishing trust within a network of entities that are otherwise unknown to one another.
And, of course, there are ethical and safety considerations – especially in a hybrid environment that inevitably brings human unpredictability. Testing of fully autonomous vehicles is accelerating the data ethics debate about how decisions are weighed up by algorithms when there is potential danger to a human being. And whose responsibility is it if an accident does occur? Is it the vehicle manufacturer, a technology provider or the public authority who authorized the vehicle? This is perhaps one of the biggest hurdles to overcome if individuals are to be willing to entrust their own lives to autonomous vehicles – algorithmic outcomes take on new significance when they may lead to driver safety taking lower priority than other 3rd parties.
What is clear for now is that it’s possible to achieve high degrees of autonomy in specific closed environments, including designated roads and tunnels, or even dedicated lanes running alongside traditional traffic.
In years to come, we can expect autonomous vehicles to operate as local swarms that make the whole experience of driving from A to B more consistent and reliable. A network of autonomous vehicles all reacting to one another’s condition and context will be able to control speeds and manage traffic flows far more efficiently, making jams and collisions a thing of the past. The challenge, in the meantime, is how to manage the transition to this safer future.
Digital Vision for Mobility
This article is part of the Atos Digital Vision for Mobility opinion paper. We explore opportunities and challenges for transport and logistics providers in this rapidly evolving space, where transport and logistics are leading other markets in digital transformation.