Using simulations in virtual reality we can increase safety and speed in experiments involving human interaction with artificial intelligence. A simulation for interaction between a pedestrian and an autonomous vehicle was developed in the CARLA simulator and Unreal Engine 4. An environment was established and equipped with the
tools necessary for facilitating interaction between the autonomous vehicle, which projects light signs onto the nearby floor, and the pedestrian. Algorithms were created to enable the autonomous vehicle to detect its surroundings and subsequently project appropriate light signs. These algorithms were developed using sensor fusion, ensuring improved data accuracy. Experimental results demonstrated the importance of clear and pre-understood signs for safe interaction. Possible upgrades to existing systems for autonomous vehicle-pedestrian interaction were proposed by integrating all three developed algorithms into one.
|
