Simulation foodchain behaviour for comparing different languages
For comparing different languages I decide to take task from book "AI Application Programming" M. Tim Jones, Charles river media, inc, 2003, Chapter 7 "Artifial life"
Modeling foodchain
Carnivore → Herbivore → Plant
- Cycled rectangled field is used that is if you take a step through bound of the field you be moved to oposite bound of the field.
- Plants is fixed object of food.
- Herbivores is moving agents that eat plants.
- Carnivores is moving agents that eat herbivores.
- Carnivores and herbivores perceive environment in its own way
- Carnivores and herbivores make decision of its action by neural network (one layer perceptron)
- If agent doesn't eat a long time, it dies
- When agent eat much enough, it can reproduce oneself, thus new agent is created in the environment.
- Each new agent has its own (random initialized) neural network that is its own new behaviour. It makes evolution process.
An areas of agent perception is defined as shown in follow table
Output signals formula
where o - output signals b - parameters of perceptron which independent from input signals u - input signals w - weight of neural network
Inputs of 1-layer perseptron are defined by next table
| index | agent | area |
|---|---|---|
| 0 | herbivore | front |
| 1 | carnivore | front |
| 2 | plant | front |
| 3 | herbivore | left |
| 4 | carnivore | left |
| 5 | plant | left |
| 6 | herbivore | right |
| 7 | carnivore | right |
| 8 | plant | right |
| 9 | herbivore | proximity |
| 10 | carnivore | proximity |
| 11 | plant | proximity |
Outputs of 1-layer perseptron are defined by next table
| index | action |
|---|---|
| 0 | turn left |
| 1 | turn right |
| 2 | move to the front |
| 3 | eat |
Executed action are selected by rule "winner take all"
- Each agent has its own level of energy which is decreased every turn and it is increased after eating
- When energy level more then particular threshold, agent can reproduces a child
- When energy level achive a zero, agent dies