Hybrid machine learning approaches for optimizing vehicle routing in moroccan urban logistics

Abstract

City vehicle routing is actually a challenge since individuals desire various things, and numerous regulations must be observed. To improve this, we developed a method that employs artificial intelligence. This new approach is a combination of several techniques: deep neural networks, reinforcement learning, and a special layer that is used to determine the optimal paths. Decisions are also made by this layer. Regulates such problems as the time spent by a driver on the road and the vehicle size. We tested this option with real data and fictitious data of a transportation company in Morocco, which has been in operation since 2012. Our results are very similar to what occurs in real life in large Moroccan cities. It considers all the regulations, and the complex trends of the time people desire things to be delivered, and the time it takes to receive them. We contrasted our approach with other general approaches. The findings indicate that our new way results in vehicles covering a shorter distance, spending less time on computers, and saving fuel. The advantage of this system is that it pre-processes most of the work, hence it can make quick decisions in real time. In general, our new system is quite effective in locating the routes of the vehicles in the Moroccan cities.