Before an object will go through their whole trajectory, we can estimate their paths, given that we know the forces that will affect them. However, when dealing with a real life problem, such as estimating a car's or a hurricane's path, most of the time we can only rely on information that describes their behaviour in the past. On the other hand, we often have data telling us the trajectory of other similar objects. In these cases, we usually can safely assume that the trajectory of the new object will be similar to those we already know of. For example, if 100 cars moving in the right lane turned right before, we can assume that the next car moving in the same lane will also take a right turn.
My task was to create an application with the help of unsupervised machine learning algorithms. This application, given information on the trajectories of previously passing cars, had to be able to recognise the trends amongst the vehicles. Then, based on the trends, it also had to estimate which lane a newly coming car would most likely take.
This project could be used as the basis for other traffic counting systems, and could also be used to predict a newly coming vehicle's trajectory. Furthermore, the algorithms used in the programme can be generalised to cluster many other objects, such as regular paths animals take, the path of various storms, and even to predict customer behaviour.