Autonomous Generation of a Public Transportation Network by an Agent-Based Model: Mutual Enrichment with Knowledge Graphs for Sustainable Urban Mobility

Abstract

Sound planning for urban mobility is a key facet of securing a sustainable future for our urban systems, and requires the careful and comprehensive assessment of its components, such as the status of the cities’ public transportation network, and how urban planners should invest in developing it. We use agent-based modelling, a tried and true method for such endeavours, for studying the history, planned future works and possible evolution of the tram line network in the Greater Geneva region. We couple these models with knowledge graphs, in a way that both are able to mutually enrich each other. Results show that the information organisation powers of knowledge graphs are highly relevant for effortlessly recounting past events and designing scenarios to be directly incorporated inside the agent-based model. The model features all 5 tram lines from the current real-world network, servicing a total of 15 communes. In turn, the model is capable of replaying past events, predicting future developments and exploring user-defined scenarios. It also harnesses its self-organisation properties to autonomously reconstruct an artificial public transportation network for the region based on two different initial networks, servicing up to 29 communes depending on the scenario. The data gathered from the simulation is effortlessly imported back into the initial knowledge graphs. The artificial networks closely resemble their real-world counterparts and demonstrate the predictive and prescriptive powers of our agent-based model. They constitute valuable assets towards a comprehensive assessment of urban mobility systems, compelling progress for the agent-based modelling field, and a convincing demonstration of its technical capabilities.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 3 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.