Connectivity Solutions to Prevent Highway Pile-ups

Cranfield University researchers in the UK are participating in a Multi-Car Collision Avoidance (MuCCA) research and development project.

This project which aims to reduce the number of motorway pile-ups involving a number of vehicles. The university says the project is examining how to prevent or reduce multi-vehicle collisions with artificial intelligence (AI) and V2V communication “to instruct autonomous vehicles to cooperatively make decisions to avoid potential incidents”.

According to Ross Walker, research fellow in autonomous cars at the university, pile-ups are rare and he suggests that motorways are the safest roads to drive along. He adds: However, when accidents do occur, they are often much more serious owing to the high speeds.  In fact, there are approximately 1,500 deaths per year, with an additional 24,000 serious injuries, and a total of 181,000 casualties of every severity.

“If a pile-up does occur then this may lead to significant loss of life, as multi-vehicle accidents account for a higher proportion of fatalities.  Therefore, helping to prevent any, no matter how rare, is a huge part of helping increase road safety.”

Charlie Wartnaby, chief engineer at Barcelona’s Institut d’Investigació Aplicada de l’Automòbil (IDIADA), adds: “Our partners at the Connected Places Catapult, which used to be the Transport Systems Catapult, in the MuCCA project wrote a report on where it would go in production. So, 2.5% of accidents on motorways or dual carriageways involve at least 2 vehicles. We haven’t a definition for a pile-up. So, it’s a small number overall. Although it’s a small proportion, high speed and multi-vehicle are likely to be disproportionate in terms of injuries and fatalities. The other societal aspect is that it can cause enormous hold-ups.”

MuCCA consortium

MuCCA is a consortium project of six organizations led by Applus+ IDIADA whose focus was on the real-time cooperation achieved in the test vehicles, while Cranfield’s key contributions were the human driver behavior model and advanced control and dynamics simulations. The university says the project was funded by Innovate UK and the Centre for Connected and Autonomous Vehicles (CCAV).  Other parties include Westfield Sports Cars, Cosworth, SBD Automotive and Connected Places Catapult.

V2X communications

As well as deploying AI, there is a role for vehicle-to-everything (V2X) communications. It can be used to alert drivers in advance about an incident ahead of them. With more vehicles connected and communicating with each other, it becomes possible for each one vehicle to relay their perception of the world to one another.

Walker adds: “If all vehicles on the road can communicate with one another in real time, then they can relay information about each one’s location and avoid any potential collision in a synchronized ‘swarm’.  This would mean that each vehicle would not need to rely on being able to visibly see the accident, as a shared ‘world view’ can be distributed between each connected vehicle.”

He finds that with V2X communications, working in real-time, each vehicle would not need to rely on being able to visibly see the accident. Together the vehicles are able to create a shared view that can be distributed between each connected or autonomous vehicle.

Even human-driven vehicles can benefit from V2X communications. The real-time information and advanced warnings they provide on visual displays can instruct each driver about what lies ahead of them, allowing them to make better driving decisions before they come upon an incident or a potentially dangerous situation. “Informing surrounding drivers of their vicinity to the accident but also, perhaps giving ‘advice’ on which lane to move to and what distance to keep, based on their prediction of human driver behaviors too,” suggests Walker.

Cameras and LiDARS

Wartnaby comments that cameras and LiDARS can be used at intersections and junctions: “Systems where the infrastructure may have sensors that track vehicles and broadcast to the vehicles, so that they are more aware of other vehicles coming towards you that you can’t see.”

At present he says there is no standardized way of doing this yet owing to this being at the research stage. However, he explains that there is also the Green Light Optimized Speed Advisory.  He says this gives details of what speed you should be going before a green light. “That’s mainly to make driving smoother and more efficient, but it may also reduce accidents to a degree as braking for lights is avoided.”

Computer algorithms

So how can computer algorithms be used to enable connected and autonomous vehicles to behave and react in “a more human-like way when avoiding collisions” and allow “any potential accidents to be recognized in advance?  Walker says it can be achieved by using machine learning techniques. For the MuCCA project this involved neural networks.

He explains: “These networks are powerful tools for learning and modelling non-linear data, which is exactly what constitutes human behaviors.  A machine learning algorithm can look at countless examples of how humans behave whilst driving and learn to recognize the patterns in those trajectories.

“By observing and learning from enough data the network can infer what a human driver may do next.  This can allow autonomous vehicles to react in human-like ways, by implementing collision avoidance trajectories similar to what the network has inferred a human driver would do in the same circumstances.  The same can be said for predicting potential accidents, as the network would also be able to predict when future trajectories of vehicles may overlap and collide.”

Wartnaby finds that nothing works without algorithms. They are key to detailing how the vehicles behave with each other. Part of the goals of the MuCCA project is about getting the vehicles to co-operate with each other – including vehicles that have at least some autonomous driving and perception capabilities to enable them to take actions autonomically and autonomously.

“If you’ve already got those things, it’s not a hardware cost feature. So, it is aimed at future vehicles with some level of autonomy, even if they aren’t fully autonomous,” he explains before adding: “We build a mathematical function, which maps the cost of taking different driving actions, steering in different directions, to avoid going off the road or colliding with other vehicles. The key thing with MuCCA is considering where other vehicles are planning to go. You can decide whether a car wants to go to the same side as you for example, and so the co-operative aspect involves deciding to go to another side to avoid a collision, so that they can get out of trouble. This works will multiple vehicles, allowing them to decide where to go. There is nobody in charge, no leader vehicle. They are all trying to help each other to achieve co-operation without any one car taking unnecessary risk itself.”

Computer simulations

Computer simulations are also proving vital for modelling how human drivers behave on motorways, helping researchers within the MuCCA project to understand the proximity of other vehicles influences behavior. This work is being carried out by Cranfield University. Walker therefore describes simulations as being essential. “They allow us to test and evaluate dangerous situations in complete safety.  It also means that tests can be performed more frequently, and much more cost-effectively, helping us to develop better systems much faster,” he comments.

With regards to vehicle proximity and how it influences behavior, Walker found that counter-intuitively having more vehicles surrounding each other actually made their behavior more predictable. He explains: “Simply because when there is less free space for each driver to move within, they are somewhat forced to move in a certain direction as they are more limited with their available options.

“For instance, with only one moving vehicle it can move almost anywhere it wanted, as it only needed to avoid hitting the accident and did not need to avoid other road users simultaneously. The real challenge with only one vehicle was successfully predicting what side of the accident it would move too and how soon, or whether it would simply break as hard as possible.”

Funding needed

He says there are a thousand more things to do, involving much of everything that’s already been done. This is because there is a need to ensure that the system is successful and reliable. There is a need for more vehicles to have V2X communications. He also wants to see more testing of collision avoidance scenarios, and there is a need to conduct testing at higher speeds. This all is done in prototype track testing, and before the challenge of integration into production vehicles. So, there is still much do and Wartnaby adds that there is also a need for more funding.

Rachel Maclean MP, Minister of Transport, Department of Transport, adds: “The potential of self-driving vehicle technology is unprecedented and could help to level up transport across the nation by making everyday journeys greener, safer, more flexible and more reliable. The MuCCA project is yet further proof of the UK leading the way in the safe and secure development of self-driving vehicle technology.”

Leave a comment

Your email address will not be published. Required fields are marked *