advoard / advoard_localization

ROS localization with uwb, odom and lidar using kalman filter method

ROS ADVOARD AGV Localization

In this project we aimed to develop a system that works in ROS environment and can localize itself. These packages tested under turtlebot3, but it can also be used on any other system that has an Ultra-wideband ranging sensor and odometry sensors mounted. It is also possible to initialize the robot and send the initial pose estimation to navigation stack without the need of manually initializing it on the RViz window but keep in mind that in order to the use this feature LiDAR is required. AMCL also needs an initial pose so this feature is very crucial to have a fully autonomous system. We offer a simulation package with synthetic UWB data for now, but it can be adjusted so that it will be possible to the use it on real life robot with real UWB sensors.

Setup

First you will need to utilize at least 4 UWB sensors one is mounted on the robot and rest should be placed in a way that it will cover the area of interest where you want to localize your robot. Simulation is possible with the shared scripts. However, this package doesn’t support direct use so slight adjustments on the pozyx scripts are required. So install the package ros_pozyx_simulation and run it before

• pozyx uwb similation
  
• pozyx uwb real (coming soon)

rosrun pozyx_simulation uwb_simulation.py rostopic echo /uwb_data_topic

After you confirm that uwb_data_topic is available in ROS you can use this project

How to Run Localization

This project includes 2 different localization algorithms and they use different sensory information.

• Kalman Filter = ultra wide band + odometry
• Source Localization = ultra wide band

Kalman Filter

rosrun advoard_localization kalman_filter_localization.py

Source Localization

rosrun advoard_localization sqrrange_leastsqr_localization.py

How to Run Initial Pose Estimation

There are 2 different initial pose estimation algorithms in this repository. First one uses LiDAR measurement, preconstructed map and UWB range measurements. Second one only uses UWB sensor and the vehicle must move 20cm forward to figure out the direction it is facing.

(Lidar + Map + UWB Sensor) Initialization

This algorithm uses uwb sensors and odometry so localization_data_topic and odom topics must to be up and running.

roslaunch advoard_localization lidar_uwb_initial_pose.launch

(UWB Sensor) Initialization

This algorithm uses only uwb sensors so localization_data_topic topic must to be up and running.

roslaunch advoard_localization uwb_initial_pose.launch

How to Use Localization Algorithm

Kalman filter and source localization use same topic thus you must select the one that you want to run because only one of them can run. The topic that mentioned is Pose. In python you can use this topic with writing a subscriber like below.

from geometry_msgs.msg import Pose

rospy.Subscriber("localization_data_topic", Pose, subscribe_data)

def subscribe_data(self,Pose): robot_realtime_pose= Pose

advoard_localization's Issues

If I can't get the gps or gazebo_modelstaes/pose

I want to set the Tag on the UAV, so I have to change the 'gazebo_modestate' to /mavros/local_postion/pose. But the pose is given from GPS. If I can't get the pose,I can't use the pozyx_simulation. So this model is to get the pose,but we have to get the pose first.Then use it to get distance, and get the pose. This is a question?

Generation of SLAM map

Dear Team,
First of all, great work on this!

Few questions on this..

1. How to generate a map and give it as an input to the algorithm.

2. How the UWB coordinate system and SLAM map coordinate system are matched exactly and how do we get the final output.

Thanks in advance

initial pose

Hi,
I have a question related to initial pose using uwb. Why we need to move robot 20 cm foward when we have the distances from anchors and position of anchors. If I only use kalman filter and set the initial pose (0,0,0), after numbers of loop, do the robot estimate its position.
Thanks

Changes in Noetic

I tried the package in Noetic, and it works brilliantly.

It needs two changes, though, to work in my setup.

• Change the shebang line to:
  #!/usr/bin/env python3

• Change the name of the message package to:
  from pozyx_simulation.msg import uwb_data

And install Scipy:
pip3 install scipy

from ieu_agv.msg import uwb_data

Thank you for your sharing at first. And I found I haven't the ieu_agv.msg file when I run the code. Could you share the file.