HOAP -Vision system

This project essentially aims at accurate throw of a ball in to the basket autonomously by the humanoid.
This is done by varying the trajectory by altering the arm's motion parameters.
The velocity and the angle are the parameters that decide the throw.

The basket(target) location is identified by the CMU Cam2.
The location coordinates are processed and converted into the motion parameters.

A joint data file for the control is generated in accordance with the required speed and the angle.

The humanoid interface is done in the Real time Linux platform.The generated joint data file is directly in the pulse counter format that is in compliance with the humanoid's firmware that controls the servo motors.

The throw is being monitored by the CMU Cam2 and the landing position of the ball is determined.The error is determined in terms of difference in distance of the ball and target form the robot. A 2 dimensional PI controller makes the necessary variations to the throw parameters and the next throw is initiated.

To overcome the shortcoming of the conventional PI controller an "Adaptive feature " is incorporated in the form of a memory database.This is done using MySQL and the throw parameters for the exact throws are being conitinuously updated in that for the future reference.

This ensures that the processing is done exactly as the human mind -"Adaptive Control", learning from the throw history.

These were practically verified using the HOAP-1 robot and the results were convincing.

Category:

Education

Tags:

• Robotics
• HOAP
• NUS
• Vision
• Education
• Controller
• Camera
• PID
• Humanoid
• basket
• basketball
• cmucam
• MSK
• Aswin
• ac
• Matlab

[Q3o5]

cool