Concept of Client-Server Environment for Agent-Based Modeling and Simulation of Living Systems

Concept of Client-Server Environment for Agent-Based Modeling and Simulation of Living Systems

Ingars Ribners and Guntis Arnicāns ([email protected], [email protected])

3 June 2015

Rīga, Latvia

Agent-Based Modeling ParadigmSystem – a composition (structure) of autonomous entities – agents.

An example of typical agent based model (Netlogo)

Modeling paradigm inspired by nature

Agent-Based Modeling Paradigm

Agent – a key abstraction in this modeling approach.

Agent means a subject that acts in some environment.

Agent-Based Modeling Paradigm

Main qualities of agent as described by Jennings et.al (1998):• Situatedness;• Autonomy;• Flexibility (responsiveness, pro-activeness, social behaviour)

Properties of Living SystemsGeneral Living Systems (GLS) theory – James G. Miller (1978, 1991)

Living System – a special subclass of open system


Eight hierarchical levels of living systems:

cells, organs, organisms, groups, organizations, communities, societies, supranational systems.


20 standard functional subsystems of a living system:

reproducer, boundary, ingestor, distributor, converter, producer, storage, extruder, motor, supporter, input transducer, internal transducer, channel and net, timer, decoder, associator, memory, decider, encoder, output transducer.

Modeling and Simulation Environment

Goal:

1. To work out a general and extensible language for describing models of systems in agent-based paradigm;

2. To work out an environment for «execution» of models described in above language that would be easy enough to be used by non-programmers.

We are using General Living Systems theory as a reference.

Features of Living Systems(that should be supported by modeling environment)

Large scale concurrent models

(up to 104-105 agents)

Comprehensive environment structures


Comprehensive environment structures

Temperature, gravitation acceleration, light, ... and their change pattern (t)

Time model

Space model

year cycle...

Day cycle,gravitation field...

Place, altitude...

An example:


Material interaction

• Environment constraints to the ability of perception;

Visibility limit


Material interaction

• Environment constraints to the ability of acting.

m𝐹→

(if m>0)

-> constraints on ,𝑣→ 𝑥→

Example 1

Example 2

msg

msg

Influence of material environmentcould be limiting as well as promoting


Non-material interaction


Individual evolution process


Memory, skills, knowledge, reasoning

Ability to discover a structure in the neighborhood


Client/server architecture

Other features of the modeling environment

• High level operations with sets;• Synchronous or asynchronous simulation;• Support of intervention into the running system;• "Avatar" feature;• Node management tools;• Openness (i.e. FIPA).

Other features of the modeling environment

Main ConceptsAgent, Event, Environment, Communication arena, Artifact






Agent structure

• A prototype of modeling environment with demo system is implemented to illustrate some of the concepts. Only wakeup functionality (once per second);

• on one Erlang/OTP 17.5 node on (PC i5-4460, 4-cores, 3.2GHz, 16GB RAM, Windows 8)

Prototype

BenchmarksNumber of

agentsStart time

(ms)*Memory

used (MB)**Memory used by agent (kB)

100 17 20,00 35,00

500 22 25,40 17,80

1000 37 32,10 15,60

5000 194 85,60 13,82

10000 530 152,40 13,59

50000 8007 726,90 14,21

100000 28451 1402,50 13,86

* Average from 4 measurements** Approx. value from Windows 8 Task Manager

Thank You!

Science

Concept of Client-Server Environment for Agent-Based Modeling and Simulation of Living Systems