project banner image

The AVHD project was designed to study how human behavior in traffic situations changes based on whether a car is controlled by another human or by an autonomous robot car.

The experiment's paradigm is simple. The participant, wearing a VR headset, is placed on one side of a street and given an incentive to reach a bus on the other side as quickly as possible without being run over. They must choose between crossing immediately, trusting that the approaching car will stop, or moving to a safe crossing point, which would constitute a detour. The experiment aims to compare decision-making when the approaching car is controlled by another participant versus when it is controlled by a computer:

graph image

Initially, I built an environment with a two-way street and cars moving at high speed:

We eventually concluded that we didn't have enough real-world space for a setup where the participant could cross a double street. Thus, we scaled it back to a single street, featuring a dense convoy of cars followed by a significant gap and then the target car. In the video below, you can see the AI I programmed for the car, represented by a green/yellow/red light. The AI takes into account the player's position and trajectory, estimating the likelihood of a collision based on several parameters:

Unfortunately, this project was shelved due to the COVID-19 pandemic, which prevented us from inviting participants and recording data. It has not been revived since.