tylerellis21 / cell-simulation

A genetic algorithm mixed with a neural network to mimic the process of evolution.

cell-simulation

A simulation that mimics the processes of nature to generate better individuates as the generations progress.

Dependencies

• sfml >= 2.3
• nex >= 0.1.0
• Glew >= 1.12.0
• Freetype >= 2.6

TODO

To be updated

Contribute

Anyone can pitch in and help contributing. Simply leave an issue or pull request. Please be consistent in coding style.

License

The MIT License

cell-simulation's Issues

Limit the amount of frames that we update the health bar.

We will gain some frame time doing this.
We also might want to skip rendering the health bar at higher zoom levels.

Add a max generation counter.

Yet another one of bombmask's crazy ideas

Use OpenCL to improve performance of cell brain activity and calculations
Will also allow the program's complexity to expand majorly.

Cell genome highscore

- Find a way to track the cell with the longest successful genome tree.

Give the cells a field of view.

Change the inputs to visual data rather than exact data.

Re-profile the simulation.

Add an view to show the graph for the current cells neural network.

Show a visualization of the currently tracked cells neural network.
Show a visual of how strong each node in the network.

Add Fancy GLSL

Don't know what, I just know that I know GLSL very well and there is some GLSL in the repo.

Leave a dead body when a cell dies.

The cell should leave something behind when it dies of natural causes.

Remove traces of the nex library.

Currently the nex library doesn't exist as it previously did.

Add the ability to let cells breed together.

Update the doxygen docs and header information.

Cmake error

CMake Error at lib/nex/CMakeLists.txt:72 (add_subdirectory):
  add_subdirectory given source
  "Z:/Source/cell-simulation/lib/nex/extlib/freetype" which is not an
  existing directory.

Convert the spatial map into a grid instead of a map.

This should allow for faster lookups.

Add some sort of "chemical markers".

Add some sort of marker that helps promote the cell breeding together.

Upgrade the genome class and implement a new trait class.

Separate the traits from the genomes since the genomes should be for the neural network only.

Implement a simulation recording option.

Somehow record the simulation output so we can rewind and view the output of each simulation. This should allow for better visual debugging and allow the user to visualize the output of the neural network.

Don't let the parents eat their own children and vise versa.

Add viewport clipping.

Don't render objects that are outside of the current view.

Add some sort of natural disaster/obstacles in the world.

This should help improve the simulation to develop some more complex interactions.

Maybe some sort of long rectangles to help split the world up.

Cell genome highscore

• Find a way to track the cell with the longest successful genome tree.
  • save most child spawns
  • possibly draw family tree to file?
• Find the most successful genome matrix and safe the top three to file or memory.
• use these high scores to implement a CELL_MINIMUM_ENTITY_COUNT to say like 50 and randomly make a highscore genome cell or a new genome cell when entity count is low

Upgrade the simulation to act more like a hive.

Maybe give the cells some sort of roll kind of like a beehive behavior.

Don't change the radius of the cell circle shape every frame.

This should also help improve the frame time a little bit.

Clean up the old doxygen format used.