SPACE (Swarm Planning And Control Evaluation) Simulator

SPACE Simulator is a pygame-based application for simulating decentralized agent behavior using behavior trees. By integrating your custom decision-making algorithms as plugins, SPACE enables rigorous testing and comparative analysis against pre-existing algorithms.

The official documentation of the SPACE simulator is available at http://space-simulator.rtfd.io/.

• Example: (Left) num_agents = 100, num_tasks = 300; (Right) num_agents = 1000, num_tasks = 3000

• Example: (Left) GRAPE; (Right) CBBA; (Common) num_agents = 30, num_tasks = 200 (static); 50 x 3 times (dynamic)

Features

• Simulates multiple agents performing tasks
• Agents use behavior trees for decision-making
• Real-time task assignment and execution
• Debug mode for visualizing agent behavior

Installation

1. Clone the repository:

   git clone https://github.com/inmo-jang/space-simulator.git
   cd space-simulator

2. Install the required dependencies:

   pip install -r requirements.txt

3. Run the simulator:

   python main.py

Configuration

Modify the config.yaml file to adjust simulation parameters:

• Number of agents and tasks
• Screen dimensions
• Agent behavior parameters

Refer to the configuration guide CONFIG_GUIDE.md

Usage

Controls

• ESC or Q: Quit the simulation
• P: Pause/unpause the simulation
• R: Start/stop recording the simulation as a GIF file

Monte Carlo Analysis

1. Set mc_runner.yaml for your purpose and run the following:

   python mc_runner.py

2. Set mc_comparison.yaml and run the following:

   python mc_analyzer.py

Code Structure

• main.py: Entry point of the simulation, initializes pygame and manages the main game loop.
• /modules/
  • agent.py: Defines the Agent class and manages agent behavior.
  • task.py: Defines the Task class and manages task behavior.
  • behavior_tree.py: Implements behavior tree nodes and execution logic.
  • utils.py: Utility functions and configuration loading.
• /plugins/
  • my_decision_making_plugin.py: Template for decision-making algorithms for each agent.

Contributing

Feel free to fork the repository and submit pull requests. Refer to TODO.md for a list of pending tasks and upcoming features.

Citations

Please cite this work in your papers!

• Inmo Jang, "SPACE: A Python-based Simulator for Evaluating Decentralized Multi-Robot Task Allocation Algorithms", arXiv:2409.04230 [cs.RO], 2024

License

GNU GPLv3