Obstacle Avoidance Behavior Acquired using Neural Q-Learning Algorithm (Validation on Actual Robot)

Researchers:
Wen Lik Dennis Lui and Velappa Ganapathy

Summary:
This video shows the acquired obstacle avoidance behavior on the Khepera II using the Neural Q-Learning Reinforcement Learning algorithm.

For more information please refer to the following papers:
1) Paper on the Neural Q-Learning Algorithm - http://www.freewebs.com/dennislui/Application%20of%20Neural%20Q-Learning%20Co...
2) Paper on the simulation environment - http://www.freewebs.com/dennislui/Application%20of%20Neural%20Q-Learning%20Co...

Category:

Science & Technology

Tags:

• Neural
• Q-Learning
• Webots
• KheperaII
• Wen
• Lik
• Dennis
• Lui
• Velappa
• Ganapathy

[wllui1]

But take note that it does not setup the radio base network. You will need to send a '*1' manually in order to connect to the robot with ID1 (but should be straightforward).

[wllui1]

Thanks. As for the radio turret, I was using a set of Matlab routines developed by Yves Piguet and Skye Legon called kMatlab. These codes are freely available and downloadable at the K-Team

website. It includes Windows DLLs to perform the system level serial port communication, and a library of useful Matlab m-files to read sensors, set speed , etc.

[arsatti]

*matlab for balbal sorry

[arsatti]

Hi there.... nice work... i can seee you are using radio turrent, which i wan unable to use (see my videos. iam using the cable). Do you know any library for radio turrent in batlab?...

Nice work though