Driverless cars are touted as the future of transport – but how can we ensure they are as safe as possible?
On May 7, 2016, a driver died when his vehicle was involved in a collision with a truck.
In March this year, a woman was walking her bicycle across a road in Arizona when she was hit by a car. She died later in hospital. Days later, a software engineer died when his car crashed into a roadside barrier and caught fire.
Three fatal crashes. Three victims. Three families grieving.
And one other similarity. One that has ensured these tragedies have made headlines and caught the attention of automotive and technology experts around the world.
All three crashes involved a vehicle that was either driverless or in self-driving mode.
These tragic crashes have brought into sharp focus the safety risks of vehicles that do not have a human driver actively monitoring the environment or reacting accordingly. Although in some of the cases the full circumstances surrounding the collision remains uncertain.
The safety of driverless cars is an issue that is worrying those in the industry. Although in these cases is a key priority for Coventry University’s Institute of Future Transport and Cities (FTC).
“These deaths have highlighted people’s fears about driverless cars. This technology is being pushed out very quickly, but we have to slow down and be sensible – it has to be safe,” said Professor Andrew Parkes, the centre’s Academic Director.
“There’s a lot of important work going on in the UK on this topic – and we’re involved in some huge collaborative projects and developing test facilities.
“Far too many people are too optimistic about these systems. We have to make sure they are tested rigorously and there are proper regulations to control them.”
One of the major problems regarding the safety of autonomous cars is the number of different situations vehicles may encounter during even just a short drive.
“There are an almost infinite number of scenarios that can happen when you’re driving. A child could run out into the road; a sudden rain shower could transform driving conditions – anything can happen,” says Dr Stratis Kanarachos, a reader in dynamics and structures at FTC.
“In order to make autonomous vehicles safe, the number of scenarios we have to test is enormous. Real life testing of all of them is just not possible.
“We have to reduce this number of scenarios and we have to bring together advanced mathematical modelling and real-world data to ensure that these automated vehicles are safer than if humans were driving them. We’re using systems engineering to facilitate this.”
The research team working on one particular project – known as Trusted Intelligent Connected Autonomous Vehicle (TIC-IT) – hope their way of deriving and modelling these scenarios and testing through a combination of test track trials and simulation will become an international standard.
“There will be an international impact to this work,” adds Dr Kanarachos, “and once we have these simulated models of how automated vehicles act in particular scenarios, we will be able to build further trust in their operation.”
The TIC-IT consortium is being led by HORIBA MIRA, in partnership with Coventry University, and was announced by the government in October last year and started research in January.
It will include the UK’s first new high speed driverless car testing track which is planned for HORIBA MIRA’s proving ground, near Nuneaton.
Another FTC project involves a Digital CAV Proving Ground that will give researchers – collaborating with automotive companies – the opportunity to test how different vehicle subsystems interact and respond to particular scenarios.
It means they can assess the hardware of an autonomous vehicle but in a simulated environment.
The key is to identify quickly the cases in which the vehicle subsystems fail or malfunction before they are implemented on a vehicle and hit the road.
”We can look at the performance of the system when driving at night or how the vehicle adjusts the driving if there’s a sudden change in the weather conditions, or the road quality, “ adds Dr Kanarachos.
“With your normal car, the assumption is that the driver is in control. In an automated car, that’s not the case. The safety of the vehicle has to be tested and regulated even further.
“The problem with automated driving is that it’s based on artificial intelligence, and in every way we have to test how it will behave in unforeseen scenarios. Artificial Intelligence is heavily dependent on the data used during the training phase. Due to the fact that there are no prescribed rules to follow the AI behaviour to a certain extent is unpredictable.”
The fact that Coventry has access to this testing system and is involved in this major project means the university is a world leader in this field.
There are already some cars on the road with low levels of automation for driving at low speeds or in traffic congestion.
There’s even been a test of driverless vehicles on city centre streets in Coventry, right through the middle of the university campus – although there were drivers in the cars to take over control if necessary.
But there are still a lot of complex issues to explore and agreements needed on safety regulations.
“This is an opportunity for us as a university, our partners and the UK to make a difference in the safety of how these vehicles are rolled out,” adds Dr Kanarachos.
“Over the next few years there will be a lot of developments in this field, and we’ll be at the heart of it.”