This project connects a trained neural network model with the Udacity Self-Driving Car Simulator to control a car autonomously based on real-time image data.

• model/model.h5: Trained Keras model file.
• drive.py: Main Python server that receives simulator images, predicts steering angles, and sends control commands back.

Follow these steps carefully:

• Download the simulator:
  • Udacity Simulator for Windows/Mac/Linux
• Install and open the simulator.
• Select "Autonomous Mode" when you start it (after you connect this server).

git clone https://github.com/sanjaybarla/self_driving_car_udacity.git
cd self_driving_car_udacity

Make sure you have Python 3.6 or higher.

pip install -r requirements.txt

python server.py

The server will start listening on port 4567. You should see

connected

• Open the Udacity Self-Driving Car Simulator.
• Select "Autonomous Mode".
• Click "Select".
• The simulator will now send camera images to your server, and your server will respond with steering and throttle commands!

This project uses a trained deep learning model to predict steering angles based on preprocessed images. Throttle is adjusted dynamically based on the speed (simple formula: throttle = 1.0 - (speed / speed_limit)). Image preprocessing includes cropping, color space conversion (RGB → YUV), Gaussian blur, resizing, and normalization.

• Incoming camera images are processed with:
• Crop (remove sky and car hood)
• Convert RGB to YUV color space
• Apply Gaussian blur
• Resize to (200x66) pixels
• Normalize pixel values between 0 and 1

This matches the Nvidia end-to-end self-driving car model training approach.