Investigating the Impact of Self-Driving Cars through Controlled Collisions

In municipalities where self-driving car testing is allowed, crossing the street without being detected by an autonomous vehicle may be a familiar experience. However, in London, this experience is not readily available. To address this, some Brits have turned it into a game called How (Not) To Get Hit By A Self-Driving Car.

The objective of the game is to cross a mocked-up crosswalk without being detected by a person-sensing camera. The game aims to demonstrate that if you can successfully make it to the other side without detection, you could also avoid being detected by an autonomous vehicle.

The creators of the game are working to make it more challenging in order to provide players with a more realistic experience. They hope that the game will allow self-driving car systems to become smarter and safer. Players who beat the AI can choose whether or not their anonymized footage will be deleted or retained by the game’s makers. These researchers are seeking research institutions with which they can share their dataset in hopes that it will be used to train self-driving cars.

Despite not knowing whether any self-driving firms use similar methods for pedestrian detection, the creators of How (Not) To Get Hit By A Self-Driving Car believe their data could still be useful. The methods that defeat this game could be reproduced by AV companies and incorporated into their own data sets or at least provide entertainment for engineers working on self-driving car technology.