An intelligent multi-agent simulation in Python where autonomous "thieves" attempt to reach a goal while evading detection by strategically placed sensors. This project explores advanced AI techniques in path planning, real-time visibility analysis, and constraint-based system validation.
🗓️ Created: May 2025 | 🎓 6th Semester, My Exploration of AI Pathfinding and Strategy
This simulation models a stealth-based scenario where multiple thieves attempt to traverse a map to reach a designated goal. They must avoid detection from a network of sensors that:
- Have a fixed radius of vision.
- Are placed such that they form a partially connected communication graph.
- Can collectively "freeze" a thief if K or more sensors simultaneously detect them.
Thieves dynamically plan their path, avoid danger zones, and react to visibility threats in real time.
- Python 3.x
- Pygame – for GUI, animation, and input
- NumPy – for matrix calculations
- Custom AI Algorithms – pathfinding and constraint validation
Before gameplay begins, the grid is constructed using constraint satisfaction logic:
- Sensor Communication CSP: Validates that at least a minimum number of sensor pairs are within communication radius.
- Placement Constraints: Ensures no overlapping entities, spatial separation, and edge-based thief placement.
- Line-of-Sight Calculation: Uses Bresenham's algorithm to detect visibility between sensors and moving agents.
- Occlusion Handling: Walls block vision; sensors cannot see through them.
Thieves don’t simply follow the shortest path. Instead, each tile in the grid is assigned a dynamic danger score based on:
- Proximity to sensors
- Visibility exposure
- Terrain type (walls, sensors, etc.)
A custom A* algorithm then navigates thieves using this weighted score map, effectively prioritizing stealth over speed.
Each thief tracks how many sensors are observing them at each step. Once detected by K or more sensors, the thief becomes permanently frozen in place, and a red line is drawn to indicate the detection event.
- Pre-Game Setup Screen: Input parameters like sensor radius, number of thieves, sensors, walls, and freeze threshold.
- Animated Simulation: Watch thieves move intelligently through the map, sensors scan zones, and the game react to every detection event.
- Real-Time Feedback: Debug outputs include visibility matrices, score maps, and thief path details.
| Component | Description |
|---|---|
| A Pathfinding* | Customized with score map to avoid danger areas |
| Score Map | Weighted grid factoring in vision overlap, proximity, and obstacles |
| Visibility Matrix | Computed per-frame using line-of-sight between sensor and thief |
| Communication Matrix | Ensures sensor network connectivity during initialization |
| Constraint Validation | CSP-style logic for ensuring a playable, fair, and valid map |