A little bit of context

A work of thesis from a collaboration between :

• The members of the Devport project who want to model ports organisations and their hinterland.
• And the LITIS whose his RI²C team (Réseaux d'Interaction et Intelligence Collective) works on complex system and on dynamic interaction networks.

What is the Seine Axis?

It is a geographical space...

Figure 1 : The Seine axis frontiers

What is the Seine Axis?

...with a huge population whose inhabitants consume goods.

Figure 2 : The population density of the Seine axis in 2007

What do we want to do?

We want to model how the goods are carried from providers to these consumers thanks to actors of logistic who must organise themselves.

What is the matter?

• At a micro level : we observe numerous and heterogeneous actors strongly interacting with each others. They organise themselves in order to build many uniques supply chains.
• At a macro level : there is an overall pattern where goods come from foreign providers and go through the Seine axis to urban areas (mainly through Le Havre and to Paris' region).
• But we can't understand the overall process only with the study of the auto-organisation of the micro level.

It is a complex system !

What will we do?

We want to use tools from the complex system theory in order to model this environment.

• Individual centred based models : multi agent systems.
• Graph theory : dynamic networks and analysis tools.

Overview

1. The Seine axis logistics.
2. Agents and graphs
3. Conceptual Model
4. The Simulation
5. The First Results

Importers and exporters

• Initiate the transportation of goods.
• Negotiate international sales contracts including the Incoterms (International Commercial Terms).

Freight forwarders

• Organise the transportation on behalf of his customer.

Transport operators

• Provide vehicles to transport goods from a place to another.
• Are responsible of goods during the transportation.

Logistic providers

• Provide a way to manage goods within and between warehouses of a supply chain.
• 4PL provide softwares and 3PL physically handle goods.

A network made of networks...

• Maritime lines.
• Road network.
• River network.
• Rail network.

... with multi-modal interfaces : the maritime and river terminals

• Allow to load (resp. unload) goods on (resp. from) a ship.
• Connected to the hinterland by others transportation mode
• Could stock temporarily goods.

Warehouses and logistics platforms

• Outsource the stock of another building.
• Could provide another logistics activities such as :
• Cross-docking.
• Container packing/unpacking.
• Quality control.
• Repackaging...

Warehouses and logistics platforms

Figure 3 : Logistics buildings in a "funnel" network (sometimes called join network)

Figure 4 : Logistics buildings in a "trumpet" network (sometimes called fork network)

What is an agent?

• It is an entity modelling either an object, an individual, a system,...
• It can have some properties which characterised it, such as : a weight, or a size, or a surface, or a shape,...
• It can have some capacities such as :
• to perceive his environment
• to manipulate his environment
• to interact with others agents

What is a graph?

What is a graph?

What is a graph?

What is a graph?

What is a graph?

Figure 6 : The road network of le Havre and his region

• An agent-based model, coupled with graphs.
• Each actor from the reality his modelled by an agent (disaggregated model).
• Generalisation of agent concept to infrastructures.
• The model can be easily coupled with other model through agents' behaviours.

Some kind of agents are actors

Figure 7 : Network representation of possible interactions between agents

Some kind of agents are actors

Figure 7a : Network representation of possible interactions between agents

Some kind of agents are actors

Figure 7b : Network representation of possible interactions between agents

Some kind of agents are actors

Figure 7c : Network representation of possible interactions between agents

Some kind of agents are actors

Figure 7d : Network representation of possible interactions between agents

Some kind of agents are infrastructures

• Some agents model physical network (road, river, ...) and have particular characteristics (max speed,...).
• Some others represent terminals. They are connected to different networks and have a surface.
• Others represent buildings such as warehouses, and they have also a surface.

Goods need to follow a determined supply chain

Figure 8 : Simplified representation of the international flow of containerised goods

The GAMA platform

The infrastructures

• Road agents : goods move on a directed and valued graph made of these Road agents, according to shortest path.
• Building agents : they represent the final consignees, and can receive and stock goods (in the limit of the free surface).
• Warehouse agents : they can stock, as Building agents, but can also send goods.

The actors

• Provider agent : it is the unique provider. It has the capacity to satisfy every order.
• FinalDestinationManager agents : theirs stocks (within a Building) are regularly consumed, so, they make some orders to restock to his LogisticProvider.
• LogisticProvider agents : they define a supply chain to each of their customers and satisfy the orders to restock. If they can not, they make an order to the Provider.

The goods

• Stock agents : they characterise the amount of goods within a Building or a Warehouse.
• Order agents : they model how many goods an actor must moved from a place to another
• Batch agents : they represent the moving goods. They have a position on the network, a speed, a destination and can have break bulks.

Analysis tools

• Charts of evolution of variables of the simulation such as : overall stock within warehouses or final destination, number of Batch agents on road...
• Neighbourhood graphs, the whole supply chains graph, collaboration graph between FinalDestinationManager and LogisticProvider.
• Computation of different measures on these graphs.

Video of the simulation

What have we done?

• Describe the logistic of the Seine axis : an environment seen as a complex system.
• Define a conceptual model based on tools often used to model complex systems.
• A multi agent system, coupled with graphs.
• Agents built from actors defined and positioned in geographical data (Shapefiles).
• Develop an agent based simulation platform in order to test and study this model.
• Develop measures of this simulation in order to improve our understanding.

What do we want to do?

• Add multi-modal network and unimplemented actors such as freight forwarder.
• Add algorithms of community detection in order to reveal stable cluster of collaborating agents.
• Confront simulation results to real data.

Thank you for your attention!