diff --git a/README.md b/README.md
index 0e7e33ee..b9c5b68a 100644
--- a/README.md
+++ b/README.md
@@ -3,20 +3,20 @@
-Evolution Gym is a large-scale benchmark for co-optimizing the design and control of soft robots. It provides a lightweight soft-body simulator wrapped with a gym-like interface for developing learning algorithms. Evogym also includes a suite of 32 locomotion and manipulation tasks, detailed on our [website](https://evolutiongym.github.io/all-tasks). Task suite evaluations are described in our [NeurIPS 2021 paper](https://arxiv.org/pdf/2201.09863).
+Evolution Gym is a large-scale benchmark for co-optimizing the design and control of soft robots. It provides a lightweight soft-body simulator wrapped with a gym-like interface for developing learning algorithms. EvoGym also includes a suite of 32 locomotion and manipulation tasks, detailed on our [website](https://evolutiongym.github.io/all-tasks). Task suite evaluations are described in our [NeurIPS 2021 paper](https://arxiv.org/pdf/2201.09863).
> [!NOTE]
-> Evogym has been recently updated! TLDR: requirements have been modernized, and the library is now pip-installable.
+> EvoGym has been recently updated! TLDR: requirements have been modernized (gym/gymnasium, numpy, etc.), and the library is now pip-installable.
[//]: # (
)
# Installation
-To use evogym, simply run:
+EvoGym supports python `3.7` to `3.10` on most operating systems:
```shell
-pip install --upgrade evogym
+pip install evogym --upgrade
```
> [!CAUTION]
@@ -26,13 +26,19 @@ pip install --upgrade evogym
> pip install -i https://test.pypi.org/simple/ evogym
> ```
+On **Linux** install the following packages (or equivalent):
+
+```shell
+sudo apt-get install xorg-dev libglu1-mesa-dev
+```
+
## From Source
-If your platform is not supported, you may try building from source:
+If your platform is not supported, you may alternatively build from source:
### Requirements
-* Python 3.7+
+* Python 3
* Linux, macOS, or Windows with [Visual Studios 2017](https://visualstudio.microsoft.com/vs/older-downloads/) build tools.
* [CMake](https://cmake.org/download/)
@@ -48,7 +54,7 @@ On **Linux only**:
sudo apt-get install xorg-dev libglu1-mesa-dev
```
-Finally, to install `evogym`, run the following command in the environment of your choice:
+Finally, to install `evogym`, run the following in the environment of your choice:
```shell
pip install -e .
@@ -88,9 +94,19 @@ if __name__ == '__main__':
This script creates a random `5x5` robot in the `Walking-v0` environment. The robot is taking random actions. A window should open with a visualization of the environment -- kill the process from the terminal to close it.
-# Usage
+## Known Issues
+### Linux and Conda
+
+Error message: `libGL error: MESA-LOADER: failed to open iris: /usr/lib/dri/iris_dri.so`
+
+Fix: `conda install -c conda-forge libstdcxx-ng`
+
+
+
+# Usage
+In addition to the resources below, you can find API documentation on our [website](https://evolutiongym.github.io/documentation).
## Tutorials
@@ -120,10 +136,6 @@ Install the necessary python requirements:
pip install -r requirements.txt
```
-## Docs
-
-You can find documentation on our [website](https://evolutiongym.github.io/documentation).
-
## Design Tool
The Design Tool provides a gui for creating Evolution Gym environments. Please see [this repo](https://github.com/EvolutionGym/evogym-design-tool).
@@ -131,6 +143,24 @@ The Design Tool provides a gui for creating Evolution Gym environments. Please s
[//]: # (
)
+## Headless Mode
+
+EvoGym runs in headless mode by default, without initializing libraries used for rendering.
+These libraries are initialized on user requests. If using a server without rendering capabilities, ensure that:
+
+```python
+# Envs are created with render_mode=None (None by default)
+env = gym.make('Walker-v0', body=body, render_mode=None)
+```
+
+```python
+# If using the low-level api, do not call EvoViewer.render()
+world = EvoWorld.from_json(os.path.join('world_data', 'simple_environment.json'))
+sim = EvoSim(world)
+viewer = EvoViewer(sim)
+viewer.render('img') # <-- Rendering libraries are initialized; do not call this
+```
+
# Dev
Install the repo with submodules:
diff --git a/examples/README.md b/examples/README.md
index 5b800545..b704a0c7 100644
--- a/examples/README.md
+++ b/examples/README.md
@@ -3,7 +3,7 @@
This readme describes how to run several control optimization and co-design experiments and visualize the results. All scripts should be run from within the `examples` directory. Ensure that you have installed requirements: `pip install -r requirements.txt` and cloned the repo with submodules before running any of these scripts.
> [!WARNING]
-> Many of these scripts have been modified in favor of improved usability. If you wish to exactly recreate the results from the original evogym paper, please see the [original release](https://github.com/EvolutionGym/evogym/releases/tag/1.0.0).
+> Many of these scripts have been modified in favor of improved usability. If you wish to exactly recreate the results from the original EvoGym paper, please see the [original release](https://github.com/EvolutionGym/evogym/releases/tag/1.0.0).
## PPO (control optimization)
@@ -39,7 +39,7 @@ python run_bo.py --eval-interval 10000 --total-timesteps 100000
python run_cppn_neat.py --eval-interval 10000 --total-timesteps 100000
```
-Note that the default parameters are set for testing purposes, and should not produce successful robots. Feel free to increase the co-design/PPO parameters based on your compute availability.
+Note that the default parameters are set for testing purposes, and will not produce task-performant robots. Feel free to increase the co-design/PPO parameters based on your compute availability. You may also reference evaluation parameters from [Appendix D. of our paper](https://arxiv.org/pdf/2201.09863).
## Visualize
diff --git a/tutorials/README.md b/tutorials/README.md
index 1cde19a0..41f03848 100644
--- a/tutorials/README.md
+++ b/tutorials/README.md
@@ -6,11 +6,11 @@ This folder contains completed code for all tutorials on our [website](https://e
To see an example of custom environment creation, see `envs/simple_env.py`. This environment is registered in `envs/__init__.py`, and can be visualized by running `python .\visualize_simple_env.py` from this directory.
-## Evogym Simulator API
+## EvoGym Simulator API
-See `basic_api.py` for a simple example of how to create, step, and render an Evogym simulator with objects of your choice. Evogym can be used to simulate any number of objects and robots (although simulation speed may slow with many objects).
+See `basic_api.py` for a simple example of how to create, step, and render an EvoGym simulator with objects of your choice. EvoGym can be used to simulate any number of objects and robots (although simulation speed may slow with many objects).
-To see understand the different rendering options available in Evogym, see `rendering_options.py`.
+To see understand the different rendering options available in EvoGym, see `rendering_options.py`.
You can run:
```bash
@@ -19,7 +19,7 @@ python .\rendering_options.py --render-option to-debug-screen
| Option | Description |
|--------------------|------------------------------------------------------------------------|
-| to-debug-screen | Render to Evogym's default viewer |
+| to-debug-screen | Render to EvoGym's default viewer |
| to-numpy-array | Render to a numpy array (visualized with open cv) |
| special-options | Render with special flags (for pretty visualization) |
| very-fast | Render without fps limit |