Developing Multi-Agent System with Natural Selection of Desire Paths for Transportation Network Planning

This research aims to propose a multi-agent model to design transportation systems that minimize the total cost in regards with transportation modes and environment topography.


Transportation systems are one of the main causes of carbon emissions and take a great part in energy consumption in terms of infrastructure, transports use and maintenance.

Their design is a complex task that requires to adopt a systemic approach to minimize their energy cost. Transport modes are not always used at their optimal speed and passenger capacity. Then, the deployment of transport modes can be reconsidered in terms of energetic efficiency. Classification of transport modes regarding their energy consumption can be made from their characteristics like optimal speed, maximal traffic rate per km or transport capacity, but environmental conditions play a high role in infrastructure design.

Overview of the research

Transportation network involves balancing various environmental (height variation, soil quality, obstacles, rivers/lake, …) and economical (maintenance, infrastructure, transports production) factors. The emergency of desire paths created as a consequence of mechanical erosion highlights misconception in relevancy and energy-efficiency of transportation systems.

Fig. 1 : Emergent desire paths from mechanical erosion of pedestrian movement

The reproduction of the emergent pathways systems and human behaviors can help civil engineers and urban planning designers to plan better transportation systems. The used multi-agent model consists of deploying a large number of agents that interact with and according to the environment to lead to a global optimized network. Pedestrian movement [1] and slime mold approximation model [2] presented first tools for solving transportation network problems. However, the relevance of infrastructure use and transport modes in terms of energy consumption are not taken into account.

An energetic approach allows to classify transport modes in terms of passenger capacity and mass transportation [4], but also according to traffic rate, lifespan of transport modes, energetic cost of transports deployment and infrastructure and maintenance.

Fig. 2 : Resulting network of 17 uniformly distributed hotspots with homogeneous attractiveness : (I) Hotspots uniform distribution, (II) Agent use frequency affordance

To evaluate this model, an image processing is applied to the resulting network and the transport modes of agent are tracked during the simulation.


[1] Ma et al., Simple agents – complex emergent path systems: Agent-based modelling of pedestrian movement Environment and Planning B: Urban Analytics and City Science, 2023

[2] Almeida and Dilão, Formation and Optimisation of Vein Networks in Physarum, 2023

[3] Schläpfer et al., The universal visitation law of human mobility, 2021

[4] Zheng et al., A universal mass-based index defining energy efficiency of different modes of passenger transport, 2021

[5] Filomena et al., Empirical characterisation of agents’ spatial behaviour in pedestrian movement simulation, 2022

[6] Hijzelendoorn, Creating Varied Terrain-Considerate Road Networks on Heightmaps with Directed Alternating Physarum Agents, Utrecht University, 2023


Louis-Andre Nicolas, 2nd year of master thesis

He entered SUPMICROTECH – ENSMM in 2019, then joined Tokyo Denki University through a double degree program in 2023.

Research Area: Cooperative Robotics, Multi-Agent Systems

  • ブックマーク