Artificial Intelligence-Driven Multimodal Route Planning: Addressing Dynamic Unavailability and Disruptions

Abstract

This paper introduces an innovative methodology for route prediction in multimodal transport networks (MTNs) utilizing Artificial Intelligence (AI). While LSTM is a well-established technique for sequence prediction, our approach uniquely leverages Monte Carlo simulations to generate a diverse and comprehensive training dataset that captures the stochastic nature of transport networks. Firstly, Monte Carlo simulations generate a comprehensive set of training samples. These samples are then processed and used to effectively train an LSTM network and predict routes upon interrogation of the obtained model. A distinctive feature of this methodology is its ability to address the issue of unavailability of the route, which could occur at any time due to unforeseen circumstances, such as weather-related disruptions. A key innovation is the introduction of a placeholder notation system within the training data to effectively model route unavailability, enabling AI to recognize and adjust to these dynamic changes adaptively. Validation through various experiments is done to demonstrate the model’s performance in accurately predicting routes, particularly in scenarios involving route unavailability. The integration of Monte Carlo simulations with LSTM networks, coupled with the unique handling of unavailable routes, marks an incremental advancement in multimodal transport route prediction and is scalable to include other variables, such as heterogeneous agent systems.

Keywords

• Distribution
• freight
• logistics
• multimodal
• route selection
• transportation