Smart Intersections & Autonomous Driving
A couple of friends and I came across some research that was being done at the University of Texas in Austin where they were exploring the possibility of automatizing vehicles entering an intersection with a scheduling approach.
Our method was really quite simple. The intersection we designed is based on Cartesian coordinates where a car travels through the intersection in each of the four direction. We set each car to begin 100 meters away and do not take into account turning or lane changes.
Rather than using a scheduling technique for occupancy of space in Cartesian coordinates for each car, we decided to parameterize the distance between each car.
Each car was given an initial velocity of 10 m/s and the objective functions were maximize velocity and minimize time. The maximum speed was set as 20 m/s, and the optimizer was left to randomly choose a car to slightly modify the acceleration of. Once the two cars’ acceleration were modified enough, all 4 cars were allowed to pass through the intersection, narrowly missing each other, but also staying 5 meters apart at all times.