Challenges:

• Variety of roads within IISc campus.
• No clear distinction / boundary between navigable and non-navigable region.
• Partial occlusion due to traffic / pedestrians.
• Other dynamic patterns like shadows etc.
• Variety of lighting conditions - morning, afternoon, evening, cloudy, sunny

Approach:

• Direct mapping from RGB images to control output using Convolutional Neural Networks (CNN).
• Simplified the output space into three control signals:
  1. Move Straight
  2. Turn Left
  3. Turn Right