Algorithm detects and avoids risks for driverless cars
Understanding how to make sure large numbers of fast-moving self-driving cars or flying robots operate in absolute safety is crucial. EU-funded researchers have figured out how, using a unique algorithm that keeps people safe from accidents or other potential disasters.
© TU Delft, 2018
The day will come when driverless vehicles are ubiquitous, possibly eliminating the human error that is involved in about 95 % of all road traffic accidents in the European Union. For some drivers, letting go of the steering wheel while cruising at 120 km/h on a highway raises concerns about safety. The EU-funded SURE project has found a solution to make sure self-driving cars dont collide as the result of faults or cyber-attacks.
By developing a unique fault-tolerant control algorithm designed to keep the vehicles at a safe distance from one another, SURE has also helped secure a future for an industry set to be worth EUR 71 billion by 2030, according to some reports. The sum corresponds to around 44 million self-driving cars.
But SUREs algorithm goes beyond self-driving cars and is adaptable and scalable for other robotic variations. It works for any scenario that entails large numbers of autonomous robots, which may for instance one day be dispatched as flying machines to map disaster areas during an emergency response situation.
SURE project coordinator Riccardo Ferrari of Delft University of Technology in the Netherlands says the threat can range from physical failure to hackers trying to cause accidents. This last category includes faulty individuals, as the result of normal or accidental breakdown, or otherwise misbehaving ones, for instance due to malicious cyber-attacks, he says.
Attack detection method
When it comes to cooperative driverless cars, it means always knowing the current speed, position and intention to accelerate or decelerate of other vehicles. The information is shared over a vehicle-to-vehicle (V2V) communication network, allowing each car to sense where it is in relation to others.
Ferrari describes SUREs solution as an attack detection method designed to protect large groups of cooperative driverless cars or other similar robot swarms. Lanes of vehicles that drive close to each other on a highway, also known as platoons, is a typical example of such a setting. Here, detecting faults and false data sent over the V2V network is paramount for safety and for avoiding collisions.
The algorithm we designed allows individual vehicles to use measurements from their on-board sensors as well as a mathematical model of the platoon dynamics to rule out false data and position themselves at a safer distance, says Ferrari.
The researchers also had to dig into the probabilities and the theory behind attacks and faults and used a novel watermarking technique to detect threats such as manipulation of data transmitted over the V2V network by hackers. The whole was simulated at a science festival in Delft last year, where the public was able to interact with a virtual two-car platoon.
From a local company to major industry
SURE is already applying its findings and is currently working with a local company in the Netherlands to help design an autonomous shuttle for passenger transport. The company used and tested approaches designed by SURE to detect any failures or false measurements.
Results are promising and we plan to further investigate this, says Ferrari, noting that they have since started to collaborate with major automotive firms. Aside from training PhD and MSc students, SURE is playing a key part in an industry that is set to revolutionise society.
SURE received funding from the EUs Marie Skłodowska-Curie Actions Individual Fellowships programme.