[Documentation] [TitleIndex] [WordIndex

  Show EOL distros: 

navigation: amcl | base_local_planner | carrot_planner | clear_costmap_recovery | costmap_2d | dwa_local_planner | fake_localization | map_server | move_base | move_base_msgs | move_slow_and_clear | nav_core | navfn | robot_pose_ekf | rotate_recovery | voxel_grid

Package Summary

The Robot Pose EKF package is used to estimate the 3D pose of a robot, based on (partial) pose measurements coming from different sources. It uses an extended Kalman filter with a 6D model (3D position and 3D orientation) to combine measurements from wheel odometry, IMU sensor and visual odometry. The basic idea is to offer loosely coupled integration with different sensors, where sensor signals are received as ROS messages.

navigation: amcl | base_local_planner | carrot_planner | clear_costmap_recovery | costmap_2d | dwa_local_planner | fake_localization | map_server | move_base | move_base_msgs | move_slow_and_clear | nav_core | navfn | robot_pose_ekf | rotate_recovery | voxel_grid

Package Summary

The Robot Pose EKF package is used to estimate the 3D pose of a robot, based on (partial) pose measurements coming from different sources. It uses an extended Kalman filter with a 6D model (3D position and 3D orientation) to combine measurements from wheel odometry, IMU sensor and visual odometry. The basic idea is to offer loosely coupled integration with different sensors, where sensor signals are received as ROS messages.

navigation: amcl | base_local_planner | carrot_planner | clear_costmap_recovery | costmap_2d | dwa_local_planner | fake_localization | map_server | move_base | move_base_msgs | move_slow_and_clear | nav_core | navfn | robot_pose_ekf | rotate_recovery | voxel_grid

Package Summary

The Robot Pose EKF package is used to estimate the 3D pose of a robot, based on (partial) pose measurements coming from different sources. It uses an extended Kalman filter with a 6D model (3D position and 3D orientation) to combine measurements from wheel odometry, IMU sensor and visual odometry. The basic idea is to offer loosely coupled integration with different sensors, where sensor signals are received as ROS messages.

navigation: amcl | base_local_planner | carrot_planner | clear_costmap_recovery | costmap_2d | dwa_local_planner | fake_localization | map_server | move_base | move_base_msgs | move_slow_and_clear | nav_core | navfn | robot_pose_ekf | rotate_recovery | voxel_grid

Package Summary

The Robot Pose EKF package is used to estimate the 3D pose of a robot, based on (partial) pose measurements coming from different sources. It uses an extended Kalman filter with a 6D model (3D position and 3D orientation) to combine measurements from wheel odometry, IMU sensor and visual odometry. The basic idea is to offer loosely coupled integration with different sensors, where sensor signals are received as ROS messages.

navigation: amcl | base_local_planner | carrot_planner | clear_costmap_recovery | costmap_2d | dwa_local_planner | fake_localization | global_planner | map_server | move_base | move_base_msgs | move_slow_and_clear | nav_core | navfn | robot_pose_ekf | rotate_recovery | voxel_grid

Package Summary

The Robot Pose EKF package is used to estimate the 3D pose of a robot, based on (partial) pose measurements coming from different sources. It uses an extended Kalman filter with a 6D model (3D position and 3D orientation) to combine measurements from wheel odometry, IMU sensor and visual odometry. The basic idea is to offer loosely coupled integration with different sensors, where sensor signals are received as ROS messages.

  • Maintainer status: maintained
  • Maintainer: David V. Lu!! <davidvlu AT gmail DOT com>, Michael Ferguson <mferguson AT fetchrobotics DOT com>
  • Author: Wim Meeussen, contradict@gmail.com
  • License: BSD
  • Source: git https://github.com/ros-planning/navigation.git (branch: hydro-devel)
navigation: amcl | base_local_planner | carrot_planner | clear_costmap_recovery | costmap_2d | dwa_local_planner | fake_localization | global_planner | map_server | move_base | move_base_msgs | move_slow_and_clear | nav_core | navfn | robot_pose_ekf | rotate_recovery | voxel_grid

Package Summary

The Robot Pose EKF package is used to estimate the 3D pose of a robot, based on (partial) pose measurements coming from different sources. It uses an extended Kalman filter with a 6D model (3D position and 3D orientation) to combine measurements from wheel odometry, IMU sensor and visual odometry. The basic idea is to offer loosely coupled integration with different sensors, where sensor signals are received as ROS messages.

  • Maintainer status: maintained
  • Maintainer: David V. Lu!! <davidvlu AT gmail DOT com>, Michael Ferguson <mfergs7 AT gmail DOT com>, Aaron Hoy <ahoy AT fetchrobotics DOT com>
  • Author: Wim Meeussen, contradict@gmail.com
  • License: BSD
  • Source: git https://github.com/ros-planning/navigation.git (branch: indigo-devel)
navigation: amcl | base_local_planner | carrot_planner | clear_costmap_recovery | costmap_2d | dwa_local_planner | fake_localization | global_planner | map_server | move_base | move_base_msgs | move_slow_and_clear | nav_core | navfn | robot_pose_ekf | rotate_recovery | voxel_grid

Package Summary

The Robot Pose EKF package is used to estimate the 3D pose of a robot, based on (partial) pose measurements coming from different sources. It uses an extended Kalman filter with a 6D model (3D position and 3D orientation) to combine measurements from wheel odometry, IMU sensor and visual odometry. The basic idea is to offer loosely coupled integration with different sensors, where sensor signals are received as ROS messages.

  • Maintainer status: maintained
  • Maintainer: David V. Lu!! <davidvlu AT gmail DOT com>, Michael Ferguson <mferguson AT fetchrobotics DOT com>
  • Author: Wim Meeussen, contradict@gmail.com
  • License: BSD
  • Source: git https://github.com/ros-planning/navigation.git (branch: jade-devel)
navigation: amcl | base_local_planner | carrot_planner | clear_costmap_recovery | costmap_2d | dwa_local_planner | fake_localization | global_planner | map_server | move_base | move_base_msgs | move_slow_and_clear | nav_core | navfn | robot_pose_ekf | rotate_recovery | voxel_grid

Package Summary

The Robot Pose EKF package is used to estimate the 3D pose of a robot, based on (partial) pose measurements coming from different sources. It uses an extended Kalman filter with a 6D model (3D position and 3D orientation) to combine measurements from wheel odometry, IMU sensor and visual odometry. The basic idea is to offer loosely coupled integration with different sensors, where sensor signals are received as ROS messages.

  • Maintainer status: maintained
  • Maintainer: David V. Lu!! <davidvlu AT gmail DOT com>, Michael Ferguson <mfergs7 AT gmail DOT com>, Aaron Hoy <ahoy AT fetchrobotics DOT com>
  • Author: Wim Meeussen, contradict@gmail.com
  • License: BSD
  • Source: git https://github.com/ros-planning/navigation.git (branch: kinetic-devel)
navigation: amcl | base_local_planner | carrot_planner | clear_costmap_recovery | costmap_2d | dwa_local_planner | fake_localization | global_planner | map_server | move_base | move_base_msgs | move_slow_and_clear | nav_core | navfn | robot_pose_ekf | rotate_recovery | voxel_grid

Package Summary

The Robot Pose EKF package is used to estimate the 3D pose of a robot, based on (partial) pose measurements coming from different sources. It uses an extended Kalman filter with a 6D model (3D position and 3D orientation) to combine measurements from wheel odometry, IMU sensor and visual odometry. The basic idea is to offer loosely coupled integration with different sensors, where sensor signals are received as ROS messages.

  • Maintainer status: maintained
  • Maintainer: David V. Lu!! <davidvlu AT gmail DOT com>, Michael Ferguson <mfergs7 AT gmail DOT com>, Aaron Hoy <ahoy AT fetchrobotics DOT com>
  • Author: Wim Meeussen, contradict@gmail.com
  • License: BSD
  • Source: git https://github.com/ros-planning/navigation.git (branch: lunar)

Package Summary

The Robot Pose EKF package is used to estimate the 3D pose of a robot, based on (partial) pose measurements coming from different sources. It uses an extended Kalman filter with a 6D model (3D position and 3D orientation) to combine measurements from wheel odometry, IMU sensor and visual odometry. The basic idea is to offer loosely coupled integration with different sensors, where sensor signals are received as ROS messages.

  • Maintainer status: unmaintained
  • Maintainer: ROS Orphaned Package Maintainers <ros-orphaned-packages AT googlegroups DOT com>
  • Author: Wim Meeussen, contradict@gmail.com
  • License: BSD
  • Source: git https://github.com/ros-planning/robot_pose_ekf.git (branch: master)

How to use the Robot Pose EKF

Configuration

A default launch file for the EKF node can be found in the robot_pose_ekf package directory. The launch file contains a number of configurable parameters:

The configuration can be modified in the launch file, which looks something like this:

Running

Nodes

robot_pose_ekf

robot_pose_ekf implements an extended Kalman filter for determining the robot pose.

Subscribed Topics

odom (nav_msgs/Odometry) imu_data (sensor_msgs/Imu) vo (nav_msgs/Odometry)

The robot_pose_ekf node does not require all three sensor sources to be available all the time. Each source gives a pose estimate and a covariance. The sources operate at different rates and with different latencies. A source can appear and disappear over time, and the node will automatically detect and use the available sensors.To add your own sensor inputs, check out the Adding a GPS sensor tutorial

Published Topics

robot_pose_ekf/odom_combined (geometry_msgs/PoseWithCovarianceStamped)

Provided tf Transforms

odom_combinedbase_footprint

How Robot Pose EKF works

Pose interpretation

All the sensor sources that send information to the filter node can have their own world reference frame, and each of these world reference frames can drift arbitrary over time. Therefore, the absolute poses sent by the different sensors cannot be compared to each other. The node uses the relative pose differences of each sensor to update the extended Kalman filter.

Covariance interpretation

As a robot moves around, the uncertainty on its pose in a world reference continues to grow larger and larger. Over time, the covariance would grow without bounds. Therefore it is not useful to publish the covariance on the pose itself, instead the sensor sources publish how the covariance changes over time, i.e. the covariance on the velocity. Note that using observations of the world (e.g. measuring the distance to a known wall) will reduce the uncertainty on the robot pose; this however is localization, not odometry.

Timing

Imagine the robot pose filter was last updated at time t_0. The node will not update the robot pose filter until at least one measurement of each sensor arrived with a timestamp later than t_0. When e.g. a message was received on the odom topic with timestamp t_1 > t_0, and on the imu_data topic with timestamp t_2 > t_1 > t_0, the filter will now update to the latest time at which information about all sensors is available, in this case to time t_1. The odom pose at t_1 is directly given, and the imu pose at t_1 is obtained by linear interpolation of the imu pose between t_0 and t_2. The robot pose filter is updated with the relative poses of the odom and imu, between t_0 and t_1.

robot_pose_ekf.png

The above figure shows experimental results when the PR2 robot started from a given initial position (green dot), driven around, and returned to the initial position. A perfect odometry x-y plot should show an exact loop closure. The blue line shows the input from the wheel odometry, with the blue dot the estimated end position. The red line shows the output of the robot_pose_ekf, which combined information of wheel odometry and imu, with the red dot the estimated end position.

Package Status

Stability

The code base of this package has been well tested and has been stable for a long time. The ROS API however has been changing as message types have evolved over time. In future versions, the ROS API is likely to change again, to a simplified single-topic interface (see Roadmap below).

Roadmap

Tutorials


2019-09-14 13:06