Turtlebot Controllers (RBE 502)

Repo Structure:

• launch/
  • ROS launch files for vicon, controllers, and simulation
• nodes/controls/
  • Controller base code
  • Specific code for PID, DFL, and NLF controllers
  • Additional code necessary for simulation
  • Post processing code
• pydoc/
  • generated html files explaning functions of the python files in nodes/controls/

Setup / Prerequisites:

• Data stream of turtlebot vicon data
• Turtlebot2 repository setup
• Gazebo for turtlebot2 setup (for simulation only)
• Controls (this repo) setup

Additionally, to use the MPC controller, the following is needed for python:

pip install cvxopt

Running:

Simulation:

To easily demonstrate the ability on simulation run the following commands in two seperate terminal windows:

roslaunch controls turtlebot.launch sim:=true
roslaunch controls controls.launch sim:=true

The 1st command will launch gazebo with the turtlbot in an empty world.

The second command will launch the controller while retrieving the fake vicon data from gazebo.

Note: the PID controller will run when no controller is specified. See below for how to specify the controller.

Actual:

In three seperate terminals run the following commands, replacing with your vicon IP address:

roslaunch controls turtlebot.launch
roslaunch controls vicon.launch ip:=<VICON IP>
roslaunch controls controls.launch

Testing other controllers (in simulation or actual):

controller.launch supports multiple arguments so please see that file for more information. But to test other controllers pre-implemented add one the following to the launch command:

ctrl:=PID
ctrl:=DFL
ctrl:=NLF
ctrl:=MPC

Note: the PID controller will run when no controller is specified.

Descriptions:

• PID is a Proportional Integral Derivative Controller
• DFL is a Dynamic Feedback Linearization Controller
• NLF is a Nonlinear Feedback Controller
• MPC is a Model Predictive Controller

Tuning Controllers:

All controller gains were obtained by tuning the controller prior to testing with terrain disturbances. The tuining procedure consisted of requesting a 0.325 m/s velocity at a 0 degree heading. The robot would start initially at a 45 degree heading or a -45 degree heading. The controller gains were changed until critical dampening of the trajectory tracking was achieved with a settling time of 2 s, which is a distance of 0.65 m from the initial position. This procedure was designed so that there was a standard metric to which all controllers were tuned to allow us to compare the transient performance of each controller with unexpected terrain disturbances.

Data Processing:

Requirements:

• rosbag the topics and run the controller
• name in the following convention controller_disturbance_trialNum.bag
  • Example: PID_board_1.bag
• be in the directory of the .bag file and ensure the directory is writable

Running:

Run the following command:

rosrun controls post_processing.py

The generated data will be saved to ./out/

Results:

https://goo.gl/5Nd63H

Additional Notes:

• Simulation tested on 16.0x : Kinetic
• Turtlebot tested on 14.0x : Indigo
• Vicon tested on 3.4.0 with 8 ceiling mounted cameras and 2 tripod mounted cameras
• Vicon data publishes to /vicon/turtlebot_traj_track/turtlebot_traj_track
• vicon IP can be set in the vicon launch arguments