Transport Secretary Chris Grayling has suggested we will see driverless cars on UK roads by 2021, but will this actually be possible? Chris Hayhurst, European Consulting Manager at MathWorks investigates.
To some extent there is already driverless technology on the road. Autonomy can be viewed on a spectrum – with level 1 having single isolated functions automated and level 5 being fully autonomous, capable of navigating any road situation without a driver. The majority of new vehicles currently on the road are level 1 or 2 with just a few claiming to be ready for level 3. However, many people don’t realise that cars are actually some of the most difficult systems to make autonomous – a consequence of the wide range of conditions and close proximity to people this form of transportation requires.
With the development and roll-out of driverless vehicles being no mean feat, it seems that we have a long way to go before we see widespread adoption. Understandably, consumer perceptions, insurance and the regulatory environment are often cited as the barriers to change - not just the technology itself. So, why are designers and engineers of autonomous vehicles finding it so challenging to realise the concept, particularly when other forms of transport are streaks ahead?
We are seeing more progress in the development of driverless vehicles off-road than on-road, as it is easier to remove people from the equation and allow for safer operation. For example, we have already seen driverless dump trucks being trialled at UK roadworks this year. There is lots more to consider when developing driverless cars for use around humans, whose movements and actions can be hard to predict, resulting in a much wider range of scenarios within which to test the driverless technology and ensure it is safe to roll-out.
One critical question is will pedestrians intentionally test the boundaries or interfere with the movements of driverless vehicles, for example by standing in front of cars so they just stop? Developers of driverless cars need to ensure they would remain safe for use around humans and continue to perform their function whether people are cooperating with the technology or not.
We are seeing better progress made in driverless vehicles required to perform repetitive tasks because they are far easier to automate, such as in agriculture with driverless tractors and mining with vehicles that haul materials from one area to another.
Heavy weather such as rain and snow make it particularly hard for driverless cars to operate, primarily because it disrupts the sensor technology used to identify signs, road markings etc. Furthermore, it is harder for driverless cars to operate safely in countries where they have poorly regulated road systems or limited road infrastructure because the driving parameters are so much more complex and unpredictable for the technology to deal with.
AI is playing a big role in the development of autonomous systems and vehicles, for example machine learning which can be used for teaching a system to recognise specific features in images or data. However, the black box issue can be a problem if the system is safety critical. The automotive industry is highly regulated, so manufacturers need to ensure their system is testable and certifiable. This can only be achieved if the AI algorithms in use can be understood by a person.
Testing is an essential part of the process when designing autonomous systems such as driverless cars. However, it’s very difficult to fully trial a system with real-world testing alone – it just isn’t possible to trial all the possible scenarios, variables and hazards that might be encountered by a vehicle in the real world. There is a real need for developers of driverless vehicles to be equipped with the right modelling technology to help them overcome this problem.
Fortunately, there is an extensive and flexible set of development tools available to engineers. These tools cover the entire development lifecycle of autonomous systems, from designing and choosing the right sensors to those that develop and deploy algorithms for vehicle perception, including simulation and control system design tools, verification and validation tools, and deployment tools.
Virtual simulation tools are particularly important as they can help recreate the myriad situations driverless vehicles may encounter in the real world. They can test for issues across a much broader range of circumstances, resulting in extensive savings in both time and money in what is a highly resource-heavy stage.
As well as simulating different road surfaces and weather patterns, it is also possible to test how a varying number of sensors, radars and cameras affect the performance of the vehicle in a simulated environment before manufacturing and real-world testing. This is particularly useful in maintaining a cost-effective programme given how expensive some sensors can be – the more sensors on a car, the more expensive it becomes, so driverless vehicle developers can use this technology to determine the trade-off.
In the future there will be two main types of autonomous features: safety features as standard and also extra convenience features like self-parking that will come at an extra cost.
In terms of AI, it will continue to be used to help develop autonomous systems, but verification and validation tools will be crucial to make sure the systems are certifiable. In addition, designers and manufacturers of autonomous vehicles need to have visibility and understanding of the decisions the AI systems are taking in order for independent third parties to certify them as safe for use around human beings.
All in all, the current likelihood of seeing fully driverless cars on the roads by 2021 is minimal, given how much testing and problem-solving there is still to be done. What is for certain is the vital role of technology in the safe and secure development of these vehicles. As investment in driverless systems grows, the road to a driverless future could accelerate.