Lane-swapping helps autonomous vehicles avoid collisions
Loughborough University has developed a computer simulation that could help developers of autonomous vehicles build in escape manoeuvres to avoid collisions with other road users.
Autonomous, driverless vehicles look set to hit the streets in the not too distant future and become increasingly common. Academics from the University’s Department of Aeronautical and Automotive Engineering have investigated algorithms that could allow such vehicles to quickly and safely switch lanes to avoid an accident.
Writing in the International Journal of Vehicle Autonomous Systems, Dr Matthew Best discusses the optimisation of a vehicle's standard brake, acceleration and steering control inputs in the context of accident prevention.
He has devised a computer simulation that shows how speed reduction and swapping lanes might be carried out by an autonomous vehicle.
The optimal rapid lane-change would inevitably be an aggressive, high ‘g’ manoeuvre that would destabilise the vehicle, and additional computing power would be needed to act quickly to correct under steer and other issues that arise during and after such a vehicle movement. The high-speed lane switch would likely be rarely used in a real-world autonomous drive, but could, in exceptional circumstances, allow driverless or robot vehicles to be safer on roads that they share with other such vehicles and vehicles with human drivers.
Dr Best points out that simulations at 70 mph (the UK national speed limit on motorways) reveal that braking alone would not lead to a safe outcome in many situations, so a lane swap would almost certainly be needed, assuming there was an empty lane for a vehicle to move into. A lane-change would in the best circumstances move the vehicle to safety in half the distance as braking at that speed.
Dr Best added: “Emergency lane change manouevres are already in consideration by motor manufacturers to help human drivers avoid accidents. This research shows the limit of performance that can be achieved when the car becomes fully autonomous.”
At present the limitations of on-board computing power in autonomous vehicles and the need for high-speed data streams means that the new algorithm is limited to simulation. However, it paves the way for developing more powerful, safety aware driving systems for such vehicles.