Machine Learning Methods for Solving Vehicle Routing Problems

The global conception of sustainable development includes a mechanism for achieving the economic goal by which all fundamental human needs are met by the use of technology with simultaneously taking care of our natural environment. To reduce air pollution and traffic we need to begin with optimizing the supply chain processes that occur on our globe every day and night. This purpose also serves every citizen by reducing everyday traffic congestions. Even minor savings produced by optimizing transportation problems are significant from the cost point of view. This paper shows recent advancements in solving the Vehicle Routing Problem (VRP) which has direct applicability in the industry. We present state-of-art machine learning and deep learning algorithms for solving VRP and Dynamic VRP where the latter makes use of simulation techniques to bring the problem near to real-world like scenarios. In the end, we propose future directions that can be further explored: a generalization of methods for arbitrary input network and the problem of multi-agent setting in non-stationary environments.