The goal of this project is to reproduce the results of the article Predator confusion is sufficient to evolve swarming behaviour
by Randal S. Olson, Arend Hintze, Fred C. Dyer, David B. Knoester, Christoph Adami
Some implementation details come from : Integrated Information Increases with Fitness in the Evolution of Animats by Jeffrey A. Edlund, Nicolas Chaumont, Arend Hintze, Christof Koch, Giulio Tononi, Christoph Adami.
Evolved preys and predator. Predator confusion reduces its chances to capture preys/
Evolved preys and predators. Agents' retina cells can be displayed
- Compile the project
- Move the rsrc folder into the application's working directory
- Launch the application with a terminal
Setting either evolve_pred or evolve_prey to true will start the evolution process.
It is necessary to disable the evolution process for evaluating the agents. Fitness values are used as file identifiers, just set predator_file_fitness_value or prey_file_fitness_value to the corresponding fitness in the saved file name to load it.
Set threads to 1 and headless to 0 to watch the agents' evolved behavior.
| Parameter | Value | Explanation |
|---|---|---|
| headless | 1 | Run the simulation without graphical display (necessary for evolving agents) |
| threads | 8 | Numbers of threads onto which to run the simulation. Each thread runs the simulation for a subset of the Markov Network Brains pool (see pool_size parameter). threads > 1 ⇔ headless 1 |
| win_w | 768 | Simulation window's horizontal size in pixels (used only when headless 0) |
| win_h | 768 | Simulation window's vertical size in pixels (used only when headless 0) |
| evolve_pred | 1 | Enable the evolution process for the predator's brain |
| evolve_prey | 1 | Enable the evolution process for the prey's brain. Setting both evolve_pred/prey to 0 enables file loading |
| prey_file_seed_value | <seed_value> |
Load the file named "prey_mb_<seed_value>.txt" storing a prey's brain |
| predator_file_seed_value | <seed_value> |
Load the file named "pred_mb_<seed_value>.txt" storing a predator's brain |
| generations | 1200 | Number of generations for evolving preys and predators |
| pool_size | 4 | Size of the brain population to evolve for both preys and predators |
| grid_h | 512 | Simulation's vertical grid size |
| grid_w | 512 | Simulation's horizontal grid size |
| ticks | 2000 | Number of steps for simulation: defines one run's duration |
| torus | 1 | Defines if the simulation takes place in a toroidal grid |
| pred/prey_fov | 120 | Predator/prey field of view in degrees |
| pred/prey_los | 100 | Predator/prey line of sight in grid units |
| pred/prey_mb_max_inputs | 4 | Maximum number of inputs predator/prey's brain's probabilistic logic gates can connect to |
| pred/prey_mb_max_outputs | 3 | Maximum number of outputs predator/prey's brain's probabilistic logic gates can connect to |
| pred/prey_mb_nb_ancestor_genes | 12 | Maximum number of outputs predator/prey's brain's |
| pred/prey_retina_cells(_by_layer) | 12 | Number of retina cells (visual inputs) for the agents. The preys have two cells' layers: one reacting to preys, the other to the predators |
| pred/prey_speed | 1 | Agents' speed in grid units |
| pred/prey_turn_speed | 8 | Agents' turnrate in degrees |
| predators/preys | 1 | Number of agents of each kind |
| predator_confusion | 1 | Enable predator's confusion |
| proba_site_copy | 25 | 1/1000th probability that a genome site get duplicated |
| proba_site_del | 50 | 1/1000th probability that a genome site get deleted |
| proba_site_insert | 25 | 1/1000th probability that a new genome site get inserted |
| proba_site_replaced | 50 | 1/1000th probability that a genome site get replaced by a new one |
| proba_site_gaussian_mutation | 50 | 1/1000th probability that a genome site undergoes a gaussian mutation |
| proba_gene_duplication | 5 | 1/1000th probability that a gene defining a PLG get duplicated |
| proba_gene_deletion | 10 | 1/1000th probability that a gene defining a PLG get deleted |
| proba_new_gene_insert | 5 | 1/1000th probability that a new gene defining a PLG get insert into the genome |
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent