In proceedings of IEEE INMIC2006 December 23-24, 2006. *
I. Mir, B.P.Amavasai and S.Meikle
Microsystems and Machine Vision Laboratory,
Materials and Engineering Research Institute,
Sheffield Hallam University, Pond Street, Sheffield S1 1WB, United Kingdom.
E-mail: mir_immad@yahoo.com
Web: https://www.shu.ac.uk/research/meri/mmvl
created with NetLogo
view/download model file: Incremental Perception V-1F.nlogo
INCREMENTAL PERCEPTION is the ability to combine information perceived by multiple agents so that the swarm as a whole may use it. Essentially, it is the phenomenon of information sharing by a swarm of un-intelligent robots without any explicit communication, which leads to a collective decision by the swarm.
Present model is our pioneer work in this field. It is a Hybrid model of Centeralised and Decenteralised architectures with homogeneous robots. These robots are un-intelligent machines and are equipped with sensors as simple as touch sensors. Like any social insects' colony, the success of the swarm lies in the cooperative behaviour of its population rather than the sophistication of individual agents. These behaviours are Movement Models, Object Search, Collision Avoidance etc. and evolve as a result of procedures like RandomHead(), HeadCarefully(), Turn(), FindObject() and LookForObject().
The swarm starts by searching an object and while doing so, the agents try to avoid collisions with each other. Once a robot touches an object, it stops moving and uses potential fields to inform other robots about the presence of an object. Every robot finding an object within a potential field will be able to deduce that at least one more robot is looking at it, and contribute to the overall perception of the object by informing the central controller, hence the term Incremental Perception.
To start with, use the following configuration:
agents = 17
signalRadius = 8
fieldCompliance = 10
robotVisionSpan = 20
movementModel = 2
object = 1
objectWeight = 10
rPower = 1
While configuring rPower and objectWeight, you might note that system behaviour is unexpected, but a close look at dependencies of the variables clearifies the situation.
Convergance time is heavily effected by swarm population, i.e. by variable 'agents'. Try using a very small value and compare it with largest possible swarm population. SignalRadius gives rise to an attractive force in which robots attract other robots. Changing this variable drastically affects the convergance time. fieldCompliance is a variable that gives rise to interesting phenomenon. Try experimenting with very small and very large values and observe the movement of robots within the field.
* In press