Visual-Guidance for Lane Keeping

Successful completion of this milestone results in:

1. Visual Features and Regions of Interest (ROI)

   We will focus primarily on regions of the image plane where yellow, white, red, and green features are expected to be found.

   1. Convert your RBG image to one of the hue structured color spaces (HSI, HSV, HSL) and use the H channel to define neighborhoods in hue space that reliably detect the four important color features.

   2. The quality, robustness, and ease of implementation all depend on the throughput of your image processing. Identify image plane regions of interest where each unique kind of feature is expected to be found. These regions should be as small as possible for processing speed and as large as necessary to make sure your robot never misses a control input.

   3. show your ROI on a representative image from duckietown.

2. Demonstrate Visually-Guided Lane Keeping
   

   devise a simple closed-loop heading controller that uses these features to track the center of a lane.

   1. Demonstrate your controller at a velocity of 10 cm/sec traversing a five tile roadway segment (3 straight sections, a left turn, and one more straight) and terminating appropriately at the end of the last straight.

   2. Demonstrate your controller at a velocity of 15 cm/sec traversing the first three (straight-away) roadway tiles and terminating at the end.

   3. reverse directions and repeat #2.1.

3. Quantify the Performance of Visually-Guided Lane Keeping
   traverse each path 5 times, compute the mean and variance of your data for each direction and report.