Agent based modeling (ABM) is an abstraction scheme suitable to capture the complexity arising from entities that dynamically interact together and take decisions based on local knowledge and external events. These are valuable characteristics when dealing with agile production processes or dynamic supply chains.
The java Enterprise Simulator (jES by Pietro Terna) is the first attempt to to investigate how enterprises arise, behave and fall, how they interact and, finally, how we can improve them using the ABM abstraction. jES follows a classic ABM approach - grounded on Swarm libraries - which is enriched with features targeting the enterprise.
It is under the vision and directions of Pietro Terna that we started to develop and improve AESOP, which is now merged with another simulation framework: Agent & Complexity in Python - ACP. The project has been founded by the University of Bologna (dep. of Management Science). The University of Torino (dep. of Economics), Fondazione Bruno Kessler (FBK) in Trento and the Div. of Packaging & Logistics of Lund University participate as a co-funding partners.
AESOP is a simulation framework aimed to production processes modeling. Compared to its ancestor - jES - it is lightweight, highly configurable, more general and it is written in Python. AESOP has been merged with ACP, which represents its new event driven core. In addition, ACP adds a network based representation of the agent relations and it is also responsible of its flexible and extensible agent configuration system. The AESOP’s scheduling mechanism - based on spreadsheet files - and the ‘recipe’ mechanism are integrated into the novel core and enriched with new features.
The basic goal of the simulator is to find bottlenecks in production process which could be hard to detect with traditional approaches (e.g., top-down). In addition, it can be used to speculate about “what-if” scenarios in order to suggest solution strategies.
Exploring organizational learning is another feature of the framework. The detection of patterns of activities leading to the emergence of routines is highlighted by the formation of a learning curve which can be plotted by the system. Learning curves would be a valuable tool for business planning, but they are hardly predictable. AESOP helps to understand the dynamic complexity of the curve formation.
A promising approach is to model and analyze the supply chain in a structured manner (network of relations). Essentially, it means showing how the structure evolves according to the actor’s behavior and external factors (e.g., market fluctuations, environmental issues, taxation growth...). This approach would help to understand the “sustainability” of a supply chain and to take action in order to improve it or mitigate a critical situation (e.g., network hubs having too much power) - this may imply socio-political consequences.
The people involved in AESOP-ACP are:
Gian Paolo Jesi, Guido Fioretti, Pietro Terna
The project is hosted at SourceForge. It can be downloaded at AESOP-ACP project page.