Installation

The recommended way to get rust is to use the rustup tool: https://rustup.rs/

cd gui

for vulkan backend (or metal backend on OS X, untested):

cargo run

If on Linux and using x11 instead of Wayland:

cargo run --fetures=x11

for other backends:

cargo run --feature="backend" --no-default-features

with "backend" replaced by "dx11", "dx12" or "gl".

Install as Python library

Install maturin (https://github.com/PyO3/maturin), then go to eco_sim_py\ffi and from within a virtualenv run

maturin develop 

Alternatively to install globally

maturin build  

Usage (GUI)

Left click to select entity, ctrl + left click to see inferred mental states for other entities, hover and tab to circle through multiple entities in the same square. Press t to toggle between sight range and threat mode for selected entity, and m to toggle showing map knowledge. Right click to instruct the seleced enityy to move to the clicked square. Space or pause button toggle the pause state, +/- adjust speed, F5 reloads the map.

Project Structure

sim contains the simulation itself, including agent behavior and mental state inference.

gui contains the GUI.

enum_macros contains helper macros used in both sim and gui.

eco_sim_py/ffi contains the Python bindings exposing sim as reinforcement learning environment, with optional use of gui for visualization.

eco_sim_py/drl uses the environment exposed in eco_sim_py/ffi and pytorch to define and train a simple NN. The NN has two heads, one predicting the rewards associated with actions, the second predicting the current mental states of other agents. The value head is used for Q-learning.