Modelling of Dynamic Configuration of Biologyinspired Multiagent Systems with Communicating Xmachine

1
Modelling of Dynamic Configuration of
Biology-inspired Multi-agent Systems with
Communicating X-machines and P Systems

• I. Stamatopoulou
• M. Gheorghe
• P. Kefalas
• South-East European Research Center,
  Thessaloniki, Greece
• (University of Sheffield, UK CITY College,
  Thessaloniki, Greece)

2
Presentation Outline

• Introduction
• Motivating Example Agent Flocking
• Communicating X-machines
• Population P systems with Active Membranes
• Conclusions

3
Introduction

• Multi-agent biology-inspired modelling
• What is an agent?
• An encapsulated computer system situated in some
  environment, capable of perceiving stimuli and
  reacting upon it inducing changes in the
  environment
• Multi-agent modelling
• Data representation (knowledge / attributes)
• Processing of data (behaviours)
• Communication

4
Introduction (cont.)

• Dynamic configuration
• Fixed system structure is not realistic
• Introduction of new individuals, growth,
  differentiation, changes in the communication
  channels between agents
• Biology-inspired systems are highly dynamic
• eg. Cell tissue, colony of ants , flock of birds
  etc.

5
Agent Flocking

• Resembles bird flocking
• Three kinds of agents
• Leaders
• Donors
• Incubators

6
Agent Flocking - Behaviours

• All types of agents
• Move freely when there is available space
  (2-dimensional)
• Avoid others by changing direction
• Have a (possibly different) perception radius
• Additionally
• Donors and Incubators follow Leaders
• Donors signal to Incubators in order to reproduce
• After mating, Donors and Incubators die and a new
  Leader is born

7
Agent Flocking (cont.)
8
Agent Communication

• Communication needs to be established
• Between a Leader and its followers (leader sends
  directional information)
• Between a Donor and an Incubator (requesting /
  accepting to mate)

9
X-machines

• An X-machine is a general computational machine
  that resembles a Finite State Machine but with
  two differences
• There is memory attached to it
• Transitions are not labelled by inputs but by
  functions that process inputs and memory values

10
Communicating X-machine System

• A Communicating X-machine System consists of
  several Communicating X-machines
• Z ((C1,, Cn), CR)
• C1,, Cn are X-machines that are able to exchange
  messages
• CR is the communication relation

11
Communicating X-machines

• Definition
• Ci (Si, Gi, Qi, Mi, FCi, Fi, q0i, m0i )
• where
• Si, Gi are the sets of inputs and outputs
• Qi is the set of states
• Mi is the memory set
• FCi is the set of functions f Si ? Mi ? Gi ? Mi
• Fi is the next state partial function
• Fi Qi ? FCi ? Qi
• q0i and m0i are the initial state and memory

12
Communicating X-machine Model (cont.)
13
Communicating X-machine Model (cont.)
14
Communicating X-machine Model (cont.)
15
Communicating X-machine Model (cont.)

• Communicating X-machine system
• Flock ((L1, I1, I2, D1, D2), (L1, I1), (L1,
  D3), (D3, I2))
• X-machines input a set of tuples of the form
• (type, position, direction)
• that describes the visible agents within the
  agents sense radius.
• Output is a set of messages.
• Memory holds the position, direction and sense
  radius.

16
Communicating X-machine Model (cont.)

• Communicating functions
• Leaders fly function
• fly (?, (pos, dir, rad))
• ((Leader, dir, pos)I1D3, (pos, dir ,
  rad)) where
• dir random(set_of_directions) and
• pos determine_pos(pos, dir)
• Donors follow_leader function
• follow_leader (Leader, L_pos, L_dirL1, (pos,
  dir, rad))
• (following_leader, (pos, dir, rad)) where
• pos determine_pos (pos, L_dir)

17
System Reconfiguration

• Application of the operators
• Generate a new component and attach it to the
  system Z using the GEN operator
• Destruct an existing component of the system Z
  using the DES operator
• Add or remove channels of communications among
  the components through the operators ATT and DET

18
System Reconfiguration (cont.)

• by a meta-level system knowing
• the Communicating System Z,
• the current system state
• definitions of the X-machine components that may
  exist in the system

19
Population P Systems with Active Membranes

• P system generalisation the structure of the
  system is an arbitrary graph
• No cells containing others
• No hierarchical organisation
• Includes operations of cell division and death

20
Population P Systems with Active Membranes (cont.)

• A Population P System with Active Membranes is a
  construct
• P (V, K, ?, a, ?E, C1, Cn, R)
• where
• V is an alphabet of symbols (objects)
• K is a finite set of cell types
• ? is a finite undirected graph
• a is a finite set of bond-making rules
• ?E is a finite multi-set of objects initially
  assigned to the environment
• each cell Ci is defined by a multi-set of objects
  it contains and its type

21
Population P Systems with Active Membranes (cont.)

• and R is a set of rules dealing with
• Communication (a b, in)t , (a b, enter)t ,
  (b, exit)t
• Transformation (a ? b)t
• Cell differentiation (a)t ? (b)s
• Cell division (a)t ? (b)t (c)t
• Cell death (a)t ?
• where a, b, c are objects and t, s are cell types.

22
Communicating X-machine Model (cont.)
23
Population P System Model

• Three types of cells K L, D, I
• The graph is (instance T)
• G (L1, I1, I2, D1, D2, I1, L1, D3, L1,
  D3, I2)
• Cells CL1 (?L1, L), CI1 (?I1, I) etc.
• ?E Ø
• Two kinds of objects
• Knowledge/attributes (position, direction etc.)
• Messages
• A bond-making rule in a might be
• (Leader, posl posd radd, Donor)
• when
• posd radd ? posl ? posd radd

24
Population P System Model (cont.)

• Sample Rules
• Communication rules
• an incubator receives seed by a donor
• (? seed, in)I
• an agent exports information to the environment
• (? pos dir, exit)L
• an agent receives percepts from the environment
• (? stimuli, enter)D
• Transformation rules
• they correspond to the agents behaviours
  (functions)
• (stimuli pos dir ? pos dir)L

25
Population P System Model (cont.)

• Sample Rules (cont.)
• Cell division rules
• these model agent birth
• (seed)I ? (toTransform)I (toDie)I
• Cell differentiation rules
• a new-born agent always becomes a Leader
• (toTransform)I ? (?)L
• Cell death rules
• (toDie)I ?

26
Problems encountered

• Modelling would be facilitated more easily if
• objects have types (my pos ? my leaders L_pos).
• constructs of objects are allowed.
• Directed output or broadcasting to the neighbours
  is allowed.

27
Problems encountered (cont.)

• Input source should is recognised in order to be
  able to engage in conversation.
• we need an option that allows the sending /
  receiving of copies of objects so that they are
  not consumed.
• environment rules that delete objects /
  information at the end of each cycle are required.

28
Conclusions

• Natural metaphor
• Performance of modelling activity
• Accuracy of the models developed
• Implications of selecting one of the two methods
  for modelling
• Similarities and possible transformation of one
  method to the other
• Tools for animation of the models