Combining Artificial Intelligence Methods to Optimize Bus Routes in a Variable Environment

Abstract

When there is a task with the freedom of many potential solutions, humans
can have a hard time finding an optimal solution in a sea of options. Computers, however,
are faster at finding solutions because they can analyze all of the available options more
rapidly than humans. This paper will be demonstrating the use of artificial intelligence
in the development of bus routes in an area. For this, we need to consider the bus stops,
the distance from one stop to the other, and the traffic in the area. Using all of that
information, we can draw the best bus route for every bus in a district. In the event
of a changing environment such as a missing bus driver or changes in traffic, a genetic
algorithm can adapt quicker to the changes than a human. In this paper, we will use
a combination of a branch and bound search and a genetic algorithm to determine the
optimal route a bus should take to pick up students in a particular area. In addition, in
the event that a machine does not provide the best route due to a lack of time for testing
alternate routes, we demonstrate that it still finds a better route than a human given the
same amount of time. Our use case demonstrates that a machine learning algorithm is
capable of performing a task with more efficiency and better results than humans alone.
These results provide evidence that machine learning using a genetic algorithm may be
used throughout the commercial industry to improve productivity in many fields.