SLAM with odometry

Hi!

Many days ago I was tried the SLAM with a simulation in Gazebo. Yesterday finally I add all nodes in my robot 4UDE!

The first setup was to setup all nodes in ROS, I’ve updated the repository with all information about robot:

https://github.com/rbonghi/ros_dude

For the first test, I’ve put my robot in the room and I’ve start to measure with 3 nodes actives:

• roslaunch dude_demo bring_up.launch rgbd_enabled:=true
This node launch this nodes:
  • Bridge to connect the unav hardware with ROS;
  • Twix-mux to combine all commands
  • RGB-D sensor to read data from x-tion sensor.
• roslaunch dude_control teleop.launch
  This node launch:
  • Interactive maker to control the robot with rviz
  • Controller to drive the robot with joystick
• roslaunch dude_navigation gmapping.launch

Robot 4UDE in the room

Finally I moved 4UDE with the joystisk and the robot start to acquire information about the environment, but to built an easy map, only one room the robot worked for more or less 20min! Amazing!

4UDE on the map

The Problem was to use only the odometry to recognize the position on the map, and I had this issues:

• Drift – With a long significant path the robot is subject to errors position
• Inaccuracy – It is more important have a fine calibration of the robot about radius and wheelbase.
• Flat plane – The robot must move in to real flat plane without unevenness and dust. The last one can change little bit the dimension of the wheel.

The map built from gmapping

With this test I decided to add another instrument to measure the orientation of the robot in the space. I’ll add an IMU!

See you at the next post! 😉

Copyright © 2015. All Rights Reserved.