Self-driving cars are expected to make road traffic safer by eliminating the “human mistake” factor. However, existing systems quickly reach their limits as some situations are difficult to predict. A research team at the Technical University of Munich has now developed an early warning system for autonomous vehicles that uses artificial intelligence (AI) to warn the driver when it foresees a situation that the system will not be able to handle on its own.
The study was carried out by a team led by Professor Eckehard Steinbach, Chair of Media Technology and member of the Board of Directors of the Munich School of Robotics and Machine Intelligence in cooperation with the BMW Group.
Instead of limiting the system to the sophisticated models that assess the behaviour of all participants on the road, the new system also relies on AI that can learn from previous situations in which self-driving vehicles have reached their limits in real-world traffic. In such scenarios, researchers claim that their system can learn from past situations where self-driving vehicles presented difficulties and so allow human to reassume control of the car.
With sensors and cameras, the system scans the environment and records status data for the vehicle including the position of the steering wheel, the condition of the road, the weather, visibility and its current speed. From these data, the AI, based on a so-called recurrent neural network, learns to recognise patterns and identify critical situations. When such situations arise, it warns the driver that a potentially critical situation is about to occur.
The technology was tested in automated driving development vehicles from the BMW Group in public road traffic. This resulted in around 2,500 situations in which drivers had to intervene. When evaluated, the study found a prediction accuracy of over 85%, with up to seven seconds of advance warning before the situations occurred.
A large amount of data is needed for the technology to work, as AI can only learn from situations it has already experienced. However, with the large number of development vehicles on the road, the data were easily generated.
Original article published by rostos.pt on 6 April 2021.
Photo Credit: © Andrey Suslov - no permission to re-use image(s) without separate licence from Shutterstock.