Applying reinforcement learning in a problem of assigning trucks to origin-destination demands under uncertainty conditions

Road transportation by truck is very vital and important in Iran. Most of the products produced in the country should to be transported by truck to be transferred to domestic markets. This paper deals with solving the problem of assigning trucks to origin-destination demands under uncertainty in travel times. Trucks are considered as intelligent agents. Then a machine learning algorithm-based on reinforcement learning is used to train the trucks. Due to the nature of the problem, a multi-agent reinforcement learning algorithm is developed and applied to solve a case study including two districts consisting of Iranian provinces. Numerical results show that the proposed approach well recognizes the appropriate allocation in each area and reduces the travel time and costs related to the transportation of trucks, and hence increases the total profitability.

Keywords:

Allocation , reinforcement learning , road transportation , truck , intelligent agent , uncertainty.