How kevin-ding-dev/cyberfly works
Cyberfly is a MuJoCo/FlyGym functional simulation for walking, sugar foraging, and maze navigation, with PPO learning and recording/evaluation workflows. The principal workflow is a user command that configures a task, runs a physical environment using a sensory or learned controller, and produces task metrics or recordings; training commands additionally evaluate and persist policies. Maze navigation adds generated layouts, odor and wall sensing, exploration memory, and dopamine-like reward signals. Edges follow shown imports, calls, or README-documented behavior; unsampled subsystem internals are represented by their exact source paths, without inferring extra call chains.
Main entry points: maze_env.py Maze environment · foraging.py Foraging environment · forage_cli.py Forage commands · maze_learning.py Maze PPO training
Components 19
Command workflows
- Command dispatcher
cyberfly/__main__.py - Forage commands
cyberfly/forage_cli.py - Maze commands
cyberfly/maze_cli.py
Simulation and sensing
- Walking environment
cyberfly/env.py - Sensory controller
cyberfly/brain.py - Foraging environment
cyberfly/foraging.py - Config validation
cyberfly/config_validation.py - Foraging checks
cyberfly/forage_checks.py
Foraging learning
- Champion challenger
cyberfly/forage_learning.py - Forage recording
cyberfly/forage_cli.py
Maze navigation
- Maze layouts
cyberfly/maze.py - Maze brain
cyberfly/maze_brain.py - Maze environment
cyberfly/maze_env.py - Maze PPO training
cyberfly/maze_learning.py - Maze checks
cyberfly/maze_checks.py
Shared learning
- PPO training and evaluation
cyberfly/learning.py - Device selection
cyberfly/training_device.py
Other components
- Researcher
- MuJoCo and FlyGym
Connections 26
- Researcher → Command dispatcher: runs command
README.md - Command dispatcher → Forage commands: dispatches forage
cyberfly/__main__.py - Command dispatcher → Maze commands: dispatches maze
cyberfly/__main__.py - Forage commands → Foraging environment: runs episodes
cyberfly/forage_cli.py - Forage commands → Champion challenger: starts training
cyberfly/forage_cli.py - Forage commands → Forage recording: records results
README.md - Foraging environment → Walking environment: extends
cyberfly/foraging.py - Foraging environment → Sensory controller: uses controller
cyberfly/foraging.py - Foraging environment → Config validation: validates config
cyberfly/foraging.py - Walking environment → MuJoCo and FlyGym: simulates with
README.md - Forage commands → Foraging checks: runs checks
README.md - Champion challenger → Foraging environment: trains and evaluates
cyberfly/forage_learning.py - Champion challenger → PPO training and evaluation: uses PPO
cyberfly/forage_learning.py - Champion challenger → Sensory controller: uses encoder
cyberfly/forage_learning.py - Maze commands → Maze PPO training: starts training
cyberfly/maze_cli.py - Maze commands → Maze environment: runs maze
cyberfly/maze_cli.py - Maze commands → Maze checks: runs checks
cyberfly/maze_cli.py - Maze environment → Foraging environment: extends
cyberfly/maze_env.py - Maze environment → Maze layouts: uses layout
cyberfly/maze_env.py - Maze environment → Maze brain: uses navigation
cyberfly/maze_env.py - Maze environment → MuJoCo and FlyGym: simulates with
README.md - Maze PPO training → Maze environment: trains and evaluates
cyberfly/maze_learning.py - Maze PPO training → Maze brain: uses encoder
cyberfly/maze_learning.py - Maze PPO training → PPO training and evaluation: uses PPO
cyberfly/maze_learning.py - Maze PPO training → Device selection: selects device
cyberfly/maze_learning.py - Maze checks → Maze environment: checks behavior
cyberfly/maze_checks.py