fmichel / madkit

MultiAgent Development Kit - version 5

MultiAgent Development Kit - MaDKit 5

MaDKit is an open source multiagent-based software written in Java.

MaDKit 5 is designed as a lightweight Java library for developing distributed applications and simulations using the multiagent paradigm.

Features

• Artificial agents creation and life cycle management
• An organizational infrastructure for communication between agents, structuring the application
• High heterogeneity in agent architectures: No predefined agent model
• Multi-Agent based simulation and simulator authoring tools
• Multi-agent based distributed application authoring facilities

Approach

In contrast to conventional approaches, which are mostly agent-centered, MaDKit follows an organization-centered approach (OCMAS): There is no predefnied agent model in MaDKit.

Especially, MaDKit does not enforce any consideration about the internal structure of agents, thus allowing a developer to freely implements its own agent architectures.

So, MaDKit is built upon the AGR (Agent/Group/Role) organizational model: Agents play roles in groups and thus create artificial societies.

Programming with MaDKit

Programming with MaDKit can be done by either :

• Downloading the zip distribution and using the provided jar file.

• Declare MaDKit as a dependency using your favorite build tool : https://mvnrepository.com/artifact/io.github.fmichel/MaDKit

For instance, with Gradle:

implementation "io.github.fmichel:MaDKit:5.x.y"

JDK 8+ is required.

More information

• Official Homepage
• API Reference
• Tutorials
• Getting started
• Documentation
• Forum

Contributing

1. Fork it
2. Create your feature branch (git checkout -b my-new-feature)
3. Commit your changes (git commit -am 'Added some feature')
4. Push to the branch (git push origin my-new-feature)
5. Create new Pull Request

madkit's Issues

Issue with thenetworking feature when virtual network interfaces are installed (e.g. docker, vbr, etc.)

MDK kernels do not connect correctly in such a case.

Multiple schedulers sharing the same simulation time

Hi,
My colleagues and I are currently developing an application on top of Madkit and we have questions about having multiple schedulers in the same simulation.

Context

For now we have a working simulation with several agents and a customized scheduler. Each agent "schedules" a bunch of actions at each tick. To do so, we created a custom ActionActivator, which is an activator that rather executing a method on all agents in a CGR location, executes an action (one activator per action).

The problem we are facing is that only the scheduler is executing the actions, decreasing performances (agents threads are idle).

If I understand well the threading model of Madkit:

• one thread per agent
• the activation of a behavior with an Activator can be distributed on several threads

The first point is fine when you have concurrent agents working at different speed, the second is designed for tick/date-based simulations. We can of course mix both.

Problem

To address our performance issue, we came up with several ideas:

1. Use the Activator#useMulticore capability. Unfortunately, given the fact that each activator executes only one action, we won't be able to distribute it.
2. Modify our action execution process: rather than scheduling actions on the scheduler, each agent manages its own actions. Then the scheduler send "top" messages to all agents, which execute scheduled actions for the current tick. This solution is good, but requires a bit of work to address agent synchronization.
3. Rather than having one scheduler, we could have multiple schedulers (up to one per regular agent). This solution could avoid us rewriting the action execution process, but is it feasible with Madkit ? I mean of course we can deploy several schedulers, but will they be synchronized ? Can they share a global simulation time ?

Thanks a lot for your time !
Pierre

Version 5.3.1 has no tag

According to website http://www.madkit.net/madkit/, current version is 5.3.1 but I can't find the related tag.
Based on commits, I would say it should be a9e6639

TickBasedTime#setCurrentTick(double) throws UnsupportedOperationException

Hi !

I have found an issue in Scheduler.TickBasedTime. I was trying to reset the simulation time in a tick-based simulation. To do so, I used Scheduler.SimulationTime#setCurrentTick(double) operation but I got an UnsupportedOperationException.

It turns out that contrary to what the documentation says, it is not a "shortcut of setCurrentTick(BigDecimal.valueOf(value))". I would have expected something like this as the body:

public default void setCurrentTick(double value){
    return setCurrentTick(BigDecimal.valueOf(value));
}

Since it is not the case, it is not enough to override setCurrentTick(BigDecimal) to have setCurrentTick(double value) working. Unfortunately, Scheduler.TickBasedTime only overrides the BigDecimal version :(

I switched back to the BigDecimal version in my code but could be great to fix it ;)

Best regards
Pierre