Driverless cars working co-operatively could cut congestion by 35%, Cambridge University researchers have said.
The estimate was given at this week's Institute of Electrical and Electronics Engineers (IEEE) 2019 International Conference on Robotics and Automation in Montréal Canada.
Working in the university's Department of Computer Science and Technology, students Yijun He and Nicholas Hyldmar fitted 16 miniature model cars with motion capture sensors and a credit card-sized Raspberry Pi computer to enable intercommunication via Wi-Fi.
They then sent these running on a two-lane track to find out what happened when one of them came to a halt. They found that the initial result when they cars moved in isolation was queues.
When, however, the vehicles' communications systems went live, these that were moving slowed down to allow them to pass the stationary one safely. When the team placed a human-controlled car on the track and drove it aggressively, the autonomous cars gave way, again ensuring safety.
The team concluded that vehicles being driven in a 'co-operative' rather than an 'egocentric' mode would improve traffic flows by 35%. Mr Hyldmar said: 'with different automotive manufacturers developing autonomous cars using their own software, the vehicles will need to communicate with each other effectively.'
The team built its own 20cm by 8cm cars using largely off-the shelf components at a cost of about £60 each, in a bid to fill the gap for low-cost platforms to support research in multi-car navigation.
Future plans include more complex scenarios, including extra lanes and intersections and a range of vehicle types.