Presentation of a logistic system

• A logistic system is part of a geographical territory composed of urban areas and of logistic structures.
• The different kind of flow (of goods, of information, or financial) are organized by actors and thanks to infrastructures.
• The goods enter and leave the system through well-known access nodes.
• Cities or urban areas attract and generate these flow.
• Different constraints (spatial, economical, political, or ecological) act over the system.

A distributed organization of the flow

Problematic

• We want to understand how actors with different goals, are organized around all the logistic infrastructures to manage flows of goods despite all the constraints of the system.
• But existing models are limited because they simulate few actors, or because they are not enough dynamic and do not allow the actors to adapt themselves to different situations.
• We propose an agent-based model which represent the properties, constraints and behaviors at a local level of a logistic system in order to reproduce the global behaviors thanks to the simulation.
• The simulation allows to test different scenarios to understand how local decisions impact the whole system.

A complex system approach

• A multi-agent model which represents each actor and infrastructure by autonomous and reactive entities.
• These agents have rules which describe how they behave and interact together according to the perception of their environment and their needs.
• Dynamic graphs represent the multi-modal transportation network.
• We can track vehicle trips in real time and observe the evolution of the traffic.

Representation of the model

The agents

How final consignees manage their collaboration with LSPs

Figure 5: How final consignees manage their collaboration with LSPs

The restock threshold

• the LSP monitors the stocks levels once a day.
• For each stock inside the warehouses of his network, the LSP determines if the current quantity of goods is too low according to this formula:
  • $q < q_\text{max} \times S$
  • where:
  • $q$ is the current quantity of goods
  • $q_\text{max}$ is the maximal quantity of goods for this stock
  • $S$ is the restock threshold.

The implementation

• We implemented the model in the GAMA simulation platform. One simulated step equals to one hour.
• We use real data about the Seine axis (around 12 000 agents) in order to validate the model and to do analysis.
• The Seine axis is mostly represented by the road (around 90% of the traffic is done by road). Therefore, we did not implement other modes of transport.

Screenshot of the simulation

Figure 7: Screenshot of the simulation

Emergence of the best restock strategies

Figure 8: Emergence of the best restock strategies: the LSPs with higher restock threshold value are more numerous at the end of the simulation.

Perturbations of the traffic

Conclusion

• We have implemented a functional simulation which uses an individual-based approach. And it is dynamic since the agents adapt themselves their behaviors.
• It allows to observe the evolution of logistic systems according to different parameters or scenarios.

Perspectives

• We are working on the implementation of the multi-modal network in order to test, for instance, the effects of the future Canal Seine-Nord project.
• We also would like to work on community detection to get a better understanding of the interactions between the agents.