Planning large systems with MDPs: case study of inland waterways supervision

• Published in: Advances in distributed computing and artificial intelligence journal Vol. 5; no. 4; pp. 71 - 84
• Main Authors: Desquesnes, Guillaume, Lozenguez, Guillaume, Doniec, Arnaud, Duviella, Éric
• Format: Journal Article
• Language: English
• Published: Salamanca Ediciones Universidad de Salamanca 01.01.2016
• Subjects: Case studies; inland waterway network; Inland waterways; large-scale systems; Markov chains; markov decision process; Model testing; Resource management; Uncertainty; Waterways
• ISSN: 2255-2863; 2255-2863
• DOI: 10.14201/ADCAIJ2016547184

Inland waterway management is likely to go through heavy changes due to an expected traffic increase in a context of climate change. Those changes will require an adaptive and resilient management of the water resource. The aim is to have an optimal plan for the distribution of the water resource on the whole inland waterway network, while taking into account the uncertainties arising from the operations of such a network. A representative model using Markov decision processes is proposed to model the dynamic and the uncertainties of the waterways. The proposed model is able to coordinate multiple entities over multiple time steps in order to prevent an overflow of a test network. However, this model suffers from a lack of scalability and is unable to represent real case applications. Advantages and limitations of several approaches of the literature to circumvent this limitation are discussed according to our case study.