Bio-entity Simulations
Jaromír Marušinec*
jaromir@marusinec.com
Abstract: This paper describes my conceptions for simulations of Biological entity in the
nature by artificial life method in 3D virtual world. Artificial live entity can be used for
simulations with live objects. Implementation possibilities are noticed in addition here.
Key Words: Artificial life, Virtual world, Virtual reality, 3D simulation.
1
Introduction
During my development of virtual reality system I discovered that this is a very good
environment for artificial life experiments. Therefore, I started my research in this theory.
I build simple objects in 3D world with motion and automatic behaviour. 2 years-research
lead to create complicated models by improving 3D objects from Virtual reality worlds. These
models control entities. Behaviour, which is very complicated and no clearly deterministic, is
called artificial life.
2
High Level Architecture of Artificial Live Entity
Sensation
Input Filter
Problem
searching
Decision
problems of
community
Problem
Priority Model
(requirement chart)
Solving
State of
feeling
Enviroment
Goal
Realization
Motion
Community
Feedback
Change of entity and
enviroment state
Fig.1: Artificial live entity architecture
Basic principles of artificial life entity simulation are
 Reactions on stimulus
 Drive for survive and growth
* Faculty of Informatics Technology, Brno University of Technology, Božetěchova 2, CZ 612 66
 Basic problems searching, decisions and solving
 State of entity and feeling
 Propagations and heredity
Fig. 2 is describes my draft of architecture of artificial live entity.
2.1
Reactions on stimulus
Input senses of artificial live entity are in virtual world unlimited. These senses can receive
any information from virtual world. For example entity can see everything and everywhere.
This is done by direct access to virtual world structures. But entity has limited capacity for
information perceptions. Therefore each of the entity has own filter for input information. See
Figure 2.
Filters
Environment
Important
sensations
Subtreshold
sensations
Entity thinking
Own model building
Fig.2: Environment sense
Entity itself never can set filter to ideal level. Sometimes happen that entity is overloaded by
senses another time it has not all information for decisions. Therefore some types of entity,
senses a minimum inputs and build own model of environment, which is important for them.
Entity, which behaviour according this model, can complete missing information from outside
and predict future. The inner model can be used for decision method, which is pretending a
situation “what will happen then...“ The decision method is own to more advance entities.
This approach is more similar to real animal and human senses. However computers in these
days have not still enough power for simulation more then one entity.
2.2
Drive for survive and growth – a priority model
First of all each entity has a priority model. A topical requirement chart is made from a static
chart by topical satisfaction. Satisfaction of requirements is not a “physical” satisfaction but it
is a “state of feeling or mind”. It means that are two values, which depends on each other.
Static chart
Feel satisfaction
Topical chart
survive
100%
warmth
drink & eat
60%
drink & eat
warmth
5%
survive
propagate
80%
propagate
explore
95%
explore
Fig.3: Personal chart making
The entity takes contents of life by satisfaction of these requirements. It mainly affects the
behaviour entity.
Implementation is done by simple sort algorithm.
2.3
Basic problems searching, decision making and experience memory
Event
stimulus
Strategic
decision making
Actions
environment
experience
Neural net
Action valuation
stimulus
Fig.4: Automatic learning scheme
This capability is based on instincts and experiences. Instincts are fixed coded reactions on
several events. These reactions can be changed in generations, but entity cannot change it in
it’s own life. On the other hand the entity get experiences that changes its behaviour in live
time. Young entity has no enough experiences for good decision. Old entity has too many
experiences, which makes their thinking slower. Therefore rejuvenation of entities is very
important.
See fig 4.
A strategic decision-making process means an instinct. This is hereditary attribute of entity
with 2 partner attributes mixing and mutations.
Neural network means database of experiences. This base is build by association an action
with valuations of this action.
Implementation of decision-making algorithm is simple model (events and reaction list),
which is used in common simulation systems. An association of an action with valuations of
this action in neural network is looks also simple. Valuation is a floating-point value it means
plus for good and minus for bad. We can ask: “But how we can describe an action and a
result generally?” Recently I am working on this problem.
2.4
State of entity and feeling
State of entity is collection of levels of entity’s power, health, energy, feeling, etc. This state
effects requirements satisfaction and capability for moving and thinking.
Environmet
Food
Energy
Force + mind
Moving and
thinking
Regeneration
Injury
Health
Feeling
Requirements
satisfaction
Fig.5: State of entity
The entity is searching mainly for food. As soon as it finds food source it is used like a
material for growing and mainly like energy for life. Energy can by used partly for
regeneration, but primary for moving and thinking. Note then in computer implementation of
thinking means consuming of processor time for decision-making.
Entity feeling is featured by many conditions, in mainly on health. This may effect
requirements satisfactions and priority chart sorting.
An implementation has problems how to store these states. I think, after experiments, that
simple solution is probably best. Every state is stored in floating point value. For example,
feeling 0.00 means normal, +1.00 means good and –1.00 is a bad feeling. Many states are in
many values, which are not stored only as a state, but also as a speed of change and
persistence.
2.5
Communication
As noted earlier, rejuvenation of entities is important. This is not only searching of partner,
and using genetic algorithms for mixing of genetic code of entity. In this context is a big
problem of communication. How to create a general language for communication?
I took an inspiration from nature again. One of my friends made a research of behaviour in his
own bee colony, for definition of bee’s language. He found more different types of messages
about bees’ life. Form and meanings of messages are special for each bee’s colony. Some
messages can be decode by any bee, but some only by bees from home colony.
My first implementation used some data token, which can be sent thought space like sound,
aroma or pheromone, gesture or dance, and data mark like excrement. For example data
mark can mean: “here is danger” or “it my territory”. Data token can mean “I’m found a
food” or “An enemy is coming”.
Now I will be experiment with XML documents with. Some information can be decoded by any
entity but all information only by entity from own colony, which ca recognise the complete
XML code.
3
Experiments
I designed several experiments in virtual graphics environment for testing these architectures.
The aim of these simulations is to create artificial individual. Who are able to find the best
way how to cope with their environment and keep their population in growing?
After that it will be possible to implement these raised entities into a real mechanic body and
use in robotics. Imagine that these robots might be a great help for firemen, rescue team,
pyrotechnics, housework, etc.
3.1
Situation
On the picture you can see several initialised entities, which are different in age, sex and
properties (random generated), and they are on the inclined plane.
There is water level on the bottom, which is rising. On the top, where is no water, are placed
food and heat sources.
3.2
Dangers
Entities can
 Drown in the water
 Freeze to death if they are far away from the heat source
 Burn down if the are too close to the heat source
 Die of hunger if they run out of their energy
 Worn-out
heat source
3.3
Goals and Basic Instincts






3.4
food
water
level
Processes





4
To drink
To eat
Be warm enough
To reproduce or breed
To survive
To explore, in case they would
satisfy all their needs.
Barriers overcome
Searching for food
Searching for heat
Reproducing
Getting more experience and skills
Entities
Fig.6: Experiment
Conclusion
Nowadays these experiments require too high computer power. If Moore law becomes true
we can expect useful results for near future.
This work is a part of The Long-term Research Program - Research in information and
control systems supported by CEZ MŠMT with code MSM 262200012.
Work is also supported by the Grant Agency of Czech Republic grants no. 102/01/1485
"Environment for Development, Modelling, and Application of Heterogeneous Systems"
References
1. Mataric, M.J., (1995) Evaluation of Learning Performance of Situated Embodied Agents in
Proceedings, Third European Conference on Artificial Life, Granada, Spain, June 4-6, 1995.
2. Adami, C., On Modelling Life Artificial Life IV: Fourth International Workshop on the Synthesis
and Simulation of Living Systems,p 269
3. Mitchell, M. (1993) Computer Models of Adaptive Complex Systems New Scientist, February 13,
1993.
4. Reidys,C. , Forst, C. V., Schuster,P. (1998) Replication and Mutation on Neutral Networks Santa
Fe Institute Working Paper 98-04-036.
5. Calabretta R., Nolfi S., and Parisi D. (1995). An Artificial life model for predicting the tertiary
structure of unknown proteins that emulate the folding process. In F. Moran, A. Moreno, J.J.
Merelo, and P Chacon (Eds.), Advances in Artificial Life: Proceedings of the Third European
Conference on Artificial Life, Berlin Heidelberg, Springer Verlag.
6. Colorni A., Dorigo, M., Maniezzo, V. (1991). Distributed Optimization by Ant Colonies.
Proceedings of ECAL91 - European Conference on Artificial Life, Paris, France, F.Varela and
P.Bourgine (Eds.), Elsevier Publishing, 134-142.
7. Grand, S., Cliff, D., Malhotra, A. (1996) Creatures: artificial life autonomous software agents for
home entertainment Cognitive Science Research Paper 434 School of Cognitive and Computing
Sciences, University of Sussex.
8. Gutowitz, H., (1994)Artificial Life Simulators and their Applications.
9. Parisi, D., Nolfi, S., Cecconi, F. (1991). Learning, Behavior and Evolution. In: Varela, F,
Bourgine, P. Towards a practice of autonomous systems. MIT Press.
Internet sources
1.
2.
3.
4.
http://www.cogs.susx.ac.uk/users/ezequiel/alife-page/alife.html
http://www.cs.brandeis.edu/~zippy/alife-library.html
http://news.alife.org
http://www.alife.net/