1 minute read

Line Tracking

Here is another example of reinforcement learning. By changing the type of reward and placing a new element in the arena, we can succeed in making the robot adopt a completely different behavior.

Load the «line tracking» configuration. Or to select the parameters yourself, start the program and then : • In AI tab: - Select «deep Q-learning» ; - In «Hidden layers» put «150 150 50» ; - Set experience buffer. • In Action tab: - put all possible actions. • In Sensor tab, select: - « camera » (for exemple 32x24)» ; - « IR sensor line tracking - signal from each of the sensors ». • In Reward tab: - Selct « Line tracking ».

- Place black tape in the arena to make a loop (if the tape is not wide, feel free to double or triple the double or triple the width of the track). - Start the robot. After a long time (about 2-3 hours) the robot will follow the track.

This article is from: