ALIEN is a CUDA-powered artificial life simulation program.
See the code
A GPU-accelerated artificial life simulator for studying emergent ecosystems
and the conditions for open-ended evolution
🏆 First prize at the ALIFE 2024 Virtual Creatures Competition
with the video Emerging Ecosystems
ALIEN (Artificial LIfe ENvironment) is an artificial life simulator built on a specialized 2D particle engine for soft bodies and fluids. Every object in a simulated world consists of particles. They can be bonded into elastic solids, flow as fluids or act as cells with specialized functions such as sensors, muscles, weapons and constructors. Each cell is controlled by its own small neural network. Networks of cells form creatures, digital organisms that perceive their environment, move, feed, compete and build offspring according to their genome. Once mutations are switched on, the simulation needs no further intervention. Populations can adapt, lineages diverge and entire ecosystems emerge.
The engine runs entirely on the graphics card and simulates worlds with millions of particles in real time. You can interact with a running world at any moment: draw matter, push creatures around, change the laws of physics or let an AI agent do it for you.
The development is driven by the desire to better understand the conditions for (pre-)biotic evolution, the growing complexity of biological systems and the open question of how evolution can keep producing novelty. At the same time, ALIEN aims to be approachable, with a modern user interface, appealing rendering and a playful spirit.
https://github.com/user-attachments/assets/fad1b719-5662-4840-aa75-165346d62cdd
ALIEN contains a built-in server for the Model Context Protocol (MCP). Connect Claude Code or any other AI agent that supports MCP and simply describe what you want, for example:
The chapter AI agents of the documentation explains the setup in detail and contains many more example prompts.
Plenty of examples can be found on the YouTube channel.
Plant-like populations around a radiation source
Close-up of different organisms showing their cell networks
Swarms attacking an ecosystem
Genome editor
| NVIDIA GPU (CUDA) | AMD GPU (HIP or SCALE) | |
|---|---|---|
| Windows | nightly build or build from sources | nightly build or build from sources |
| Linux | build from sources | build from sources |
| Cloud instance | Docker image, headless | not available |
NVIDIA: compute capability 7.5 or higher, i.e. GeForce RTX 20 series or newer (list of supported GPUs). AMD: RDNA2 or newer.
Download and unpack https://alien-project.org/files/alien-develop.zip, which is built every night from the develop branch. It contains two executables:
alien.exe for NVIDIA GPUsalien-amd.exe for AMD GPUs (RDNA2, RDNA3 and RDNA4)Start it directly from the unpacked folder, otherwise it will not find the resource folder. If the program crashes for an unknown reason, please refer to the troubleshooting section below.
For long runs without local hardware, a nightly image containing the headless command-line interface is published on Docker Hub as chrxh/alien:nightly. It holds cli and its resource folder. There is no GUI in the image.
On a rented GPU instance (for example vast.ai), enter chrxh/alien:nightly as the instance image and filter the offers for a GeForce RTX 20 to 50 series GPU. Upload your simulation file to the instance with scp, then connect via SSH and start it by hand:
cd /opt/alien
cli -i example.sim -o output.sim
Locally, with an NVIDIA GPU and the NVIDIA container toolkit installed:
docker run --rm --gpus all -v "$PWD":/data --entrypoint cli chrxh/alien:nightly -i /data/example.sim -o /data/output.sim -t 1000
Windows and Linux are built the same way, using the cross-platform CMake build system, the Ninja build tool and the vcpkg package manager, which is included as a Git submodule.
Getting the sources
Open a command prompt in a suitable directory (which should not contain whitespace characters) and enter:
git clone --recursive https://github.com/chrxh/alien.git
The --recursive parameter is necessary to check out the vcpkg submodule as well. Submodules are not updated by a plain git pull, so use git pull --recurse-submodules instead.
Prerequisites
sudo apt-get install ninja-build libx11-dev libxcursor-dev libxrandr-dev libxinerama-dev libxi-dev libxext-dev libxfixes-dev libgl1-mesa-dev libglu-dev libxcb1-dev pkg-config
NVIDIA GPUs (CUDA)
On Windows, just run build-windows-ninja.bat from the repository root. It sets up the MSVC environment and locates Ninja and CMake from your Visual Studio installation automatically.
On Linux, or on Windows from a Developer Command Prompt, invoke the CMake preset directly:
cmake --preset ninja
cmake --build --preset ninja-release
The executable is written to build-ninja/Release/ (alien.exe on Windows, alien on Linux) and has to be started from there, otherwise it will not find the resource folder.
AMD GPUs (HIP)
The HIP path builds the same sources for AMD GPUs in place of CUDA. On Windows: build-windows-ninja.bat HIP, which writes to build-ninja-hip/Release/ and covers RDNA2, RDNA3 and RDNA4.
On Linux:
cmake --preset ninja-hip -DCMAKE_HIP_ARCHITECTURES=gfx1100 -DCMAKE_PREFIX_PATH=/opt/rocm
cmake --build --preset ninja-hip-release
CMAKE_HIP_ARCHITECTURES selects the target architecture (gfx1100 for RDNA3, gfx90a for CDNA2 / MI200). If omitted, it is auto-detected from the GPUs of the host. CMAKE_PREFIX_PATH lets find_package(hip) locate the ROCm installation when CMake uses the vcpkg toolchain. Adjust it if ROCm is installed elsewhere.
AMD GPUs (SCALE)
Alternatively, the unchanged CUDA sources can be compiled for AMD GPUs with SCALE (Linux), whose nvcc is a drop-in replacement for the NVIDIA one:
cmake --preset ninja -DCMAKE_CUDA_COMPILER=/opt/scale/bin/nvcc -DCMAKE_CUDA_ARCHITECTURES=gfx1100
cmake --build --preset ninja-release
Besides the graphical program, ALIEN includes cli (cli.exe on Windows), which runs simulations without a window. It is useful for long experiments, performance measurements and the automated evaluation of simulations with different parameters. For example,
cli -i example.sim -o output.sim -t 1000
runs the simulation file example.sim for 1000 time steps and writes the result to output.sim. Without -t, the simulation runs until you press Q or Ctrl+C. All options are described in the chapter Files and command line.
ALIEN comes with a complete documentation built into the program. Press F1 to open it. It is organized in three parts:
The same chapters can also be read here on GitHub in resources/docs.
If ALIEN does not start or crashes, please make sure that:
alien.exe, click Options and choose High performance.If the error still occurs, enable Settings > Debug mode, reproduce the error and create a GitHub issue with the log.txt attached. More solutions can be found in the chapter Troubleshooting.
If you use ALIEN in a scientific publication, please cite:
Heinemann, C.: Artificial Life Environment. Informatik-Spektrum 31, 55–61 (2008). https://doi.org/10.1007/s00287-007-0205-1
@article{Heinemann2008,
author = {Heinemann, Christian},
title = {Artificial Life Environment},
journal = {Informatik-Spektrum},
volume = {31},
number = {1},
pages = {55--61},
year = {2008},
doi = {10.1007/s00287-007-0205-1}
}
Note that this paper describes the beginnings of what has become ALIEN. The current simulator differs substantially from the system presented there. You are therefore equally welcome to cite this repository instead of the paper or alongside it: https://github.com/chrxh/alien
If you adopt ideas or concepts from ALIEN without using the software itself, that is fine but please just mention ALIEN and link to this repository.
Contributions to the project are very welcome. The most convenient way is to communicate via GitHub Issues, Pull requests or the Discussion forum depending on the subject. For example, it could be
ALIEN has been initiated, mainly developed and maintained by Christian Heinemann. Many thanks to everyone who has contributed to this project in any way. In alphabetical order:
The following external libraries are used:
Free icons and icon font:
ALIEN is licensed under the BSD 3-Clause license.
(top 24 of 57)
1,314 followers · starred Aug 2024
3,457 followers · starred Aug 2024
1,558 followers · starred Jun 2021
2,245 followers · starred Nov 2024
ALIEN is a CUDA-powered artificial life simulation program.
See the code
A GPU-accelerated artificial life simulator for studying emergent ecosystems
and the conditions for open-ended evolution
🏆 First prize at the ALIFE 2024 Virtual Creatures Competition
with the video Emerging Ecosystems
ALIEN (Artificial LIfe ENvironment) is an artificial life simulator built on a specialized 2D particle engine for soft bodies and fluids. Every object in a simulated world consists of particles. They can be bonded into elastic solids, flow as fluids or act as cells with specialized functions such as sensors, muscles, weapons and constructors. Each cell is controlled by its own small neural network. Networks of cells form creatures, digital organisms that perceive their environment, move, feed, compete and build offspring according to their genome. Once mutations are switched on, the simulation needs no further intervention. Populations can adapt, lineages diverge and entire ecosystems emerge.
The engine runs entirely on the graphics card and simulates worlds with millions of particles in real time. You can interact with a running world at any moment: draw matter, push creatures around, change the laws of physics or let an AI agent do it for you.
The development is driven by the desire to better understand the conditions for (pre-)biotic evolution, the growing complexity of biological systems and the open question of how evolution can keep producing novelty. At the same time, ALIEN aims to be approachable, with a modern user interface, appealing rendering and a playful spirit.
https://github.com/user-attachments/assets/fad1b719-5662-4840-aa75-165346d62cdd
ALIEN contains a built-in server for the Model Context Protocol (MCP). Connect Claude Code or any other AI agent that supports MCP and simply describe what you want, for example:
The chapter AI agents of the documentation explains the setup in detail and contains many more example prompts.
Plenty of examples can be found on the YouTube channel.
Plant-like populations around a radiation source
Close-up of different organisms showing their cell networks
Swarms attacking an ecosystem
Genome editor
| NVIDIA GPU (CUDA) | AMD GPU (HIP or SCALE) | |
|---|---|---|
| Windows | nightly build or build from sources | nightly build or build from sources |
| Linux | build from sources | build from sources |
| Cloud instance | Docker image, headless | not available |
NVIDIA: compute capability 7.5 or higher, i.e. GeForce RTX 20 series or newer (list of supported GPUs). AMD: RDNA2 or newer.
Download and unpack https://alien-project.org/files/alien-develop.zip, which is built every night from the develop branch. It contains two executables:
alien.exe for NVIDIA GPUsalien-amd.exe for AMD GPUs (RDNA2, RDNA3 and RDNA4)Start it directly from the unpacked folder, otherwise it will not find the resource folder. If the program crashes for an unknown reason, please refer to the troubleshooting section below.
For long runs without local hardware, a nightly image containing the headless command-line interface is published on Docker Hub as chrxh/alien:nightly. It holds cli and its resource folder. There is no GUI in the image.
On a rented GPU instance (for example vast.ai), enter chrxh/alien:nightly as the instance image and filter the offers for a GeForce RTX 20 to 50 series GPU. Upload your simulation file to the instance with scp, then connect via SSH and start it by hand:
cd /opt/alien
cli -i example.sim -o output.sim
Locally, with an NVIDIA GPU and the NVIDIA container toolkit installed:
docker run --rm --gpus all -v "$PWD":/data --entrypoint cli chrxh/alien:nightly -i /data/example.sim -o /data/output.sim -t 1000
Windows and Linux are built the same way, using the cross-platform CMake build system, the Ninja build tool and the vcpkg package manager, which is included as a Git submodule.
Getting the sources
Open a command prompt in a suitable directory (which should not contain whitespace characters) and enter:
git clone --recursive https://github.com/chrxh/alien.git
The --recursive parameter is necessary to check out the vcpkg submodule as well. Submodules are not updated by a plain git pull, so use git pull --recurse-submodules instead.
Prerequisites
sudo apt-get install ninja-build libx11-dev libxcursor-dev libxrandr-dev libxinerama-dev libxi-dev libxext-dev libxfixes-dev libgl1-mesa-dev libglu-dev libxcb1-dev pkg-config
NVIDIA GPUs (CUDA)
On Windows, just run build-windows-ninja.bat from the repository root. It sets up the MSVC environment and locates Ninja and CMake from your Visual Studio installation automatically.
On Linux, or on Windows from a Developer Command Prompt, invoke the CMake preset directly:
cmake --preset ninja
cmake --build --preset ninja-release
The executable is written to build-ninja/Release/ (alien.exe on Windows, alien on Linux) and has to be started from there, otherwise it will not find the resource folder.
AMD GPUs (HIP)
The HIP path builds the same sources for AMD GPUs in place of CUDA. On Windows: build-windows-ninja.bat HIP, which writes to build-ninja-hip/Release/ and covers RDNA2, RDNA3 and RDNA4.
On Linux:
cmake --preset ninja-hip -DCMAKE_HIP_ARCHITECTURES=gfx1100 -DCMAKE_PREFIX_PATH=/opt/rocm
cmake --build --preset ninja-hip-release
CMAKE_HIP_ARCHITECTURES selects the target architecture (gfx1100 for RDNA3, gfx90a for CDNA2 / MI200). If omitted, it is auto-detected from the GPUs of the host. CMAKE_PREFIX_PATH lets find_package(hip) locate the ROCm installation when CMake uses the vcpkg toolchain. Adjust it if ROCm is installed elsewhere.
AMD GPUs (SCALE)
Alternatively, the unchanged CUDA sources can be compiled for AMD GPUs with SCALE (Linux), whose nvcc is a drop-in replacement for the NVIDIA one:
cmake --preset ninja -DCMAKE_CUDA_COMPILER=/opt/scale/bin/nvcc -DCMAKE_CUDA_ARCHITECTURES=gfx1100
cmake --build --preset ninja-release
Besides the graphical program, ALIEN includes cli (cli.exe on Windows), which runs simulations without a window. It is useful for long experiments, performance measurements and the automated evaluation of simulations with different parameters. For example,
cli -i example.sim -o output.sim -t 1000
runs the simulation file example.sim for 1000 time steps and writes the result to output.sim. Without -t, the simulation runs until you press Q or Ctrl+C. All options are described in the chapter Files and command line.
ALIEN comes with a complete documentation built into the program. Press F1 to open it. It is organized in three parts:
The same chapters can also be read here on GitHub in resources/docs.
If ALIEN does not start or crashes, please make sure that:
alien.exe, click Options and choose High performance.If the error still occurs, enable Settings > Debug mode, reproduce the error and create a GitHub issue with the log.txt attached. More solutions can be found in the chapter Troubleshooting.
If you use ALIEN in a scientific publication, please cite:
Heinemann, C.: Artificial Life Environment. Informatik-Spektrum 31, 55–61 (2008). https://doi.org/10.1007/s00287-007-0205-1
@article{Heinemann2008,
author = {Heinemann, Christian},
title = {Artificial Life Environment},
journal = {Informatik-Spektrum},
volume = {31},
number = {1},
pages = {55--61},
year = {2008},
doi = {10.1007/s00287-007-0205-1}
}
Note that this paper describes the beginnings of what has become ALIEN. The current simulator differs substantially from the system presented there. You are therefore equally welcome to cite this repository instead of the paper or alongside it: https://github.com/chrxh/alien
If you adopt ideas or concepts from ALIEN without using the software itself, that is fine but please just mention ALIEN and link to this repository.
Contributions to the project are very welcome. The most convenient way is to communicate via GitHub Issues, Pull requests or the Discussion forum depending on the subject. For example, it could be
ALIEN has been initiated, mainly developed and maintained by Christian Heinemann. Many thanks to everyone who has contributed to this project in any way. In alphabetical order:
The following external libraries are used:
Free icons and icon font:
ALIEN is licensed under the BSD 3-Clause license.
(top 24 of 57)
1,314 followers · starred Aug 2024
3,457 followers · starred Aug 2024
1,558 followers · starred Jun 2021
2,245 followers · starred Nov 2024