FireScout: EKF-SLAM Robotics Simulation

FireScout is a modular Python simulation demonstrating Simultaneous Localization and Mapping (SLAM) using an Extended Kalman Filter (EKF).

Designed as a portfolio project to showcase Object-Oriented Architecture, State Estimation, and Autonomous Navigation.

Key Features

• Extended Kalman Filter (EKF): Implements the full predict-update cycle with Jacobian linearization for landmarks.
• Modular Architecture: strictly decoupled Robot (Kinematics), Environment (Ground Truth), and EKF (Estimation) modules.
• Wall-Following Logic: Autonomous navigation controller using simulated LIDAR-like sensor data.
• Visual Analytics: Generates post-mission mapping analysis comparing Ground Truth vs. EKF Estimation.

🛠️ Project Structure

FireScout-SLAM/
├── main.ipynb            # Interactive Notebook & Simulation
├── README.md             # Documentation
└── src/                  # Source Package
    ├── config.py         # Configuration & Hyperparameters
    ├── ekf.py            # EKF Math & Matrix Operations
    ├── environment.py    # Map Generation & Feature Management
    ├── robot.py          # Robot Physics & Sensor Simulation
    └── simulation.py     # Orchestrator & Visualization

QUICK START

1. Clone the repo git clone https://github.com/the1quadfather/FireScout-SLAM.git cd FireScout-SLAM

2. Install dependencies pip install numpy matplotlib tqdm

3. Run! Open main.ipynb and run all the cells! The parmeters in Cell 2 ("Config") can be changed to suit your curiosity. python -m src.simulation

RESULTS

The simulation produces a GIF and a map showing:

• Green Line: Robot's true path
• Blue Dots: EKF's estimated position (should track very closely to the green line)
• Purple X: EKF's estimated landmark positions (fire, hazards, etc.)
• Orange & Red Icons: Actual landmarks

THEORY EXPLAINED

• Motion model is based on a differntial drive's kinematics with Gaussian noise.
• Observation model is a range-bearing feature extraction method
• Landmark association is done via nearest-neighbor feature matching