Automated cars are designed by people



For autonomous technology to educate itself it needs raw real world data. The problem is that very little actual accident data is available or can be simulated. Near miss incidents are far more frequent though and it is exactly the data gathered from these higher frequency occurrences that enable cars with Dreams4Cars technology to “dream” better reactions to unexperienced situations.

Researchers and car makers have been developing autonomous vehicles for years now; designing software to operate within known parameters and requiring extensive field tests to ensure high safety standards. Such tests require vehicles drive 10.000-15.000 miles a week in order to collect enough data to test the software, and for each new release data logs of 3.000.000 km.

Despite all the great progress made and the millions of test miles driven we are still someway off having adequate data for the kind of software update necessary for autonomous systems to anticipate all possible events, particularly rare ones. A properly designed cognitive simulation technology is needed with the capacity to learn by itself from the data it has collected and generate more scenarios and as a result produce more data to analyse and process around every critical scenario. This learning by “dreaming” will therefore enable cars to foresee events their designers did not anticipate and will significantly speed up the development of autonomous vehicles and get them ever closer to the threshold of being safer than human drivers. This is what Dreams4Cars is about.

Go to the overview of "Cars Dream"

This project has received funding from the European Union's Horizon 2020 research and innovation programme under grant agreement No 731593.