28shambhavi / arm_mpc Goto Github PK

This ROS package implements a Model Predictive Controller on a two DOF planar robot arm. The setup can be used for implementing other controllers as well. Calculation of the dynamic model is based on kinetic and potential energies. Euler-Lagrange equations are used to compute the equations of motion. The final dynamic model depends on the arm's moment of inertia, coriolis force and gravitation force. [2] The nonlinear dynamic model is multivariable (with two control inputs and two controlled outputs) so a two step design is proposed where we first develop a feedback linearisation control and then design the controller. [3]

Build package from source: navigate to the source folder of your catkin workspace and build this package using:

$ git clone https://github.com/28shambhavi/arm_mpc.git
$ cd ..
$ catkin_make

1. Arm model used can be changed by modifying the urdf in urdf/2linkrobot.urdf. It is defined as base>arm1>arm2>endEff.

2. Navigate into the launch folder. Launch the desired controller in the rviz environment.

$ roscd arm_mpc/launch
$ roslaunch control_display.launch

3. Using the controllers: The launch file named display.launch launches the forward kinematics controller from scripts/controller.py for any point in the arm's workspace. The second launch file, control_display.launch will launch the MPC controller from scripts/controller_mpc.py and show the path from an initial point (both angles making zero degrees to the x axis) to the desired location of the end effector.

1. Python3
2. ROS Noetic

1. Murray, Richard; A Mathematical Introduction to Robotic Manipulation
2. David I, Robles G. PID Control Dynamics of A Robotics Arm Manipulator with Two Degrees of Freedom.Control De Procesos y Robótica. 2012, pp. 1–7.
3. Guechi E-H, Bouzoualegh S, Zennir Y, Blažič S. MPC Control and LQ Optimal Control of A Two-Link Robot Arm: A Comparative Study. Machines. 2018; 6(3):37. https://doi.org/10.3390/machines6030037

Shambhavi Singh ( GitHub, Website )