Objectives

This is one of the assignments in the subject 41013 Industrial Robotics that I am really proud of sharing with everyone. The requirements for this project are:

  • Creating a custom robot arm with custom DH parameters and design that serve for a specific purpose.
  • Using the RMRC to plan the path for the robot.
  • Integrating the collision avoiance and collision detection with safety environment.
  • Design the GUI with teach function to control the robot, there must be an Estop in the GUI to stop the robot immediately if something is out of control.
  • Integrating the second robot that provided in the toolbox to do the task.

    • Why did I choose a redundant 7 DOF robot?

  • Increase of the flexibility for hitting the shuttlecock
  • Improve of the manipulability: Easier to avoid the singularities
  • Task optimization: The collision avoidance can be ultilized, enabling the robot find the most efficient way to complete the task.

    • Motion of the shuttlecock

  • The motion of the shuttlecock can be simplified into 1 scenario, for example, there is a high-speed camera that can detect the velocity and the angle of the shuttlecock , and I assume that the user can input these values, the initial velocity and the angle.
  • The projectile motion formula is applied to create 3 matrices of x, y and z coordinates. The size of one of them is 100x1
  • As I assumed there are no air friction or wind, the x coordinate of it will be the same which is 0, at the middle of the court.

    • Where should it hit?

    • Firstly, I find the position of the endeffector by taking the forward kinematic with the function fkine()
    • Then, I create a zeros matrix of 100x1 for storing the distance.
    • I take the distance in the Y direction of the endeffector to each point on the trajectory of the shuttlecock that I have already predicted before.
    • Then I find the mininum this distance of it and which point will give the least distance.

    Image Based Visual Seroving

    I also integrate the visual servoing for the robot, this method call image based visual servoing, this means the camera mounted on the endeffector and we can map the velocity of the image pixel to the velocity of the endeffector by multiply the image Jacobian matrix to the velocity of the endeffector. The video on the right demonstrates how the image based visual seroving work, I mount a camera on the endeffector but I make it invisible, the desired frame is that camera will be at the center of the opponent racket at 500x500px.

    Final video demonstration

    This is the final video demonstration for the system, it can demonstrate better how I apply the RMRC and path planning between poses. The method that I used to detect the collision is ellipsoid check, it means creating an ellipsoid around the links of the robot and check the distance between the points and the radius of the ellipsoid to see if it lie on the surface or inside or outside the surface. RMRC is integrated in generating the hitting trajectory of the shuttlecock, after finding the hitting point, racket will move backward a bit just like when we swing the racket to have more inertia, and then it will move in a straight line with RMRC to move to the hitting point. All the poses are generated with inverse kinematic to find the joint states for the robot. All the code can be found on my GitHub.

    Conclusion

  • After finishing this project with HD, I have learnt many interesting knowledge in industrial robotics:
  • Inverse and foward kinematic, trajectory planning between poses with trapezodial and quintic polynomial velocity profile.
  • Collision check and collision avoidance.
  • Resolved motion rate control and visual servoing method.
  • Design a safety environment to work with robot.
  • OOP programming in MATLAB and design a GUI for robot in MATLAB, integrate the Arduino with Matlab to control the gripper for the Dobot.
  • ROS toolbox in MATLAB to control the Dobot.