V-REP Simulation from ELEC3210

People

Name	Email	Task
Jiekai ZHANG	jzhanger@connect.ust.hk	1, 2, 3, 6
Yutian ZHANG	yzhangkq@connect.ust.hk	4, 5

Before starting

1. Make sure your ROS interface of CoppeliaSim (or V-REP) is working fine
2. There are some modification to the env.ttt, these are necessary for the project to run normally
   • Rename ROSInterface to ROS in the script of the bot
   • Changed the endPos in the script of the moving ball's path
3. Install the following ros package
   • joy
   • teleop_twist_keyboard
   • hector mapping
   • rviz
   • opencv and bridge
   • (gmapping, robot_localization)

Tasks and method

Please refer to the report for more detail

Task 1, Bulid a 2D map

Using hector_mapping (robot_map)

Done, but the performance is not quite well. It works only when the bot moves very slowly.
But it do not require a odom frame, so very simple.
Also a wired thing is the orientation of map frame and base_link is not aligned. I just added a frame to fix it but there should have a better way.

Using gmapping (robot_nav)

Cannot get a very good odom using robot_localization so cannot use the map.
May be try to integrate the odom by myself (should be much better by my test).
P.S. There's no odom provide by the simulator and the control command /vrep/vel_cmd is not usable since the real speed in the simulator does NOT match the input speed. I really wonder why there is a such stupid mistake.

Task 2, Use keyboard to control the bot (robot_controller)

Done, using teleop_twist_keyboard for keyboard and joy for joy con.
A unsatisfying fact is that the teleop_twist_keyboard will mess up the terminal output, looks very ugly. Maybe try turtlebot_teleop later. Currently I only use joy con for controlling.

Task 3, Image localization with recognition (face_detection, utils)

Done.
The first step is extract the image on the wall using find counters. (Or you can try extrace the face directly using Haar cascades in OpenCV)
Then using eigen face to recognition it.
The final step is to calculate the angle of the image, then get the distance from laser scan.

Task 4, Follow the sphere (sphere_tracking)

Done, filter the yellow color directly and you will get a perfect circle. Apply find counter to it and calculate its error.

Task 5, Localizaiton (robot_position)

Done, hardcord the distance, nothing interesting

Task 6, Launch file (robot_bringup)

Done