Changelog for package imu_filter_madgwick

1.1.8 (2020-05-25)

• Drop the signals component of Boost (#103) Unused and is removed in Boost 1.71 (Ubuntu Focal). Signals2 is header-only.
• Add the option to remove the gravity vector (#101)
• Rewrite rosbags: Use MagneticField for magnetometer
• fix install path & boost linkage issues
• Contributors: Alexis Paques, Martin Günther, Mike Purvis, Sean Yen

1.1.7 (2019-05-06)

• Skip messages and warn if computeOrientation fails
• Contributors: Martin Günther

1.1.6 (2018-05-25)

• Remove outdated Makefile
• update to use non deprecated pluginlib macro (#77)
• Add warning when IMU time stamp is zero Closes #82.
• Contributors: Martin Günther, Mikael Arguedas

1.1.4 (2017-05-22)

• Print warning if waiting for topic Closes #61.
• Fix boost::lock_error on shutdown
• Contributors: Martin Günther

1.1.3 (2017-03-10)

• Return precisely normalized quaternions Fixes #67 : TF_DENORMALIZED_QUATERNION warning added in TF2 0.5.14.
• Tests: Check that output quaternions are normalized
• Fixed lock so it stays in scope until end of method.
• Contributors: Jason Mercer, Martin Guenther, Martin Günther

1.1.2 (2016-09-07)

• Add missing dependency on tf2_geometry_msgs
• Contributors: Martin Guenther

1.1.1 (2016-09-07)

• Add parameter "world_frame": optionally use ENU or NED instead of NWU convention (from #60; closes #36)
• Add parameter "stateless" for debugging purposes: don't do any stateful filtering, but instead publish the orientation directly computed from the latest accelerometer (+ optionally magnetometer) readings alone
• Replace the (buggy) Euler-angle-based initialization routine (ImuFilterRos::computeRPY) by a correct transformation matrix based one (StatelessOrientation::computeOrientation) and make it available as a library function
• Refactor madgwickAHRSupdate() (pull out some functions, remove micro optimizations to improve readability)
• Add unit tests
• Contributors: Martin Guenther, Michael Stoll

1.1.0 (2016-04-25)

1.0.11 (2016-04-22)

• Jade: Change default: use_magnetic_field_msg = true
• Contributors: Martin Guenther

1.0.10 (2016-04-22)

1.0.9 (2015-10-16)

1.0.8 (2015-10-07)

1.0.7 (2015-10-07)

1.0.6 (2015-10-06)

• Split ImuFilter class into ImuFilter and ImuFilterRos in order to have a C++ API to the Madgwick algorithm
• Properly install header files.
• Contributors: Martin Günther, Michael Stoll

1.0.5 (2015-06-24)

• Add "~use_magnetic_field_msg" param. This allows the user to subscribe to the /imu/mag topic as a sensor_msgs/MagneticField rather than a geometry_msgs/Vector3Stamped. The default for now is false, which preserves the legacy behaviour via a separate subscriber which converts Vector3Stamped to MagneticField and republishes.
• Contributors: Mike Purvis, Martin Günther

1.0.4 (2015-05-06)

• update dynamic reconfigure param descriptions
• only advertise debug topics if they are used
• allow remapping of the whole imu namespace with this change, all topics can be remapped at once, like this: rosrun imu_filter_madgwick imu_filter_node imu:=my_imu
• Contributors: Martin Günther

1.0.3 (2015-01-29)

• Add std dev parameter to orientation estimate covariance matrix
• Port imu_filter_madgwick to tf2
• Switch to smart pointer
• Contributors: Paul Bovbel, Martin Günther

1.0.2 (2015-01-27)

• fix tf publishing (switch parent + child frames) The orientation is between a fixed inertial frame (fixed_frame_) and the frame that the IMU is mounted in (imu_frame_). Also, imu_msg.header.frame should be imu_frame_, but the corresponding TF goes from fixed_frame_ to imu_frame_. This commit fixes that; for the reverse_tf case, it was already correct. Also see http://answers.ros.org/question/50870/what-frame-is-sensor_msgsimuorientation-relative-to/. Note that tf publishing should be enabled for debug purposes only, since we can only provide the orientation, not the translation.
• Add ~reverse_tf parameter for the robots which does not have IMU on root-link
• Log mag bias on startup to assist with debugging.
• add boost depends to CMakeLists All non-catkin things that we expose in our headers should be added to the DEPENDS, so that packages which depend on our package will also automatically link against it.
• Contributors: Martin Günther, Mike Purvis, Ryohei Ueda

1.0.1 (2014-12-10)

• add me as maintainer to package.xml
• turn mag_bias into a dynamic reconfigure param Also rename mag_bias/x --> mag_bias_x etc., since dynamic reconfigure doesn't allow slashes.
• gain and zeta already set via dynamic_reconfigure Reading the params explicitly is not necessary. Instead, dynamic_reconfigure will read them and set them as soon as we call config_server->setCallback().
• reconfigure server: use proper namespace Before, the reconfigure server used the private namespace of the nodelet manager instead of the nodelet, so the params on the parameter server and the ones from dynamic_reconfigure were out of sync.
• check for NaNs in magnetometer message Some magnetometer drivers (e.g. phidgets_drivers) output NaNs, which is a valid way of saying that this measurement is invalid. During initialization, we simply wait for the first valid message, assuming there will be one soon.
• magnetometer msg check: isnan() -> !isfinite() This catches both inf and NaN. Not sure whether sending inf in a Vector3 message is valid (Nan is), but this doesn't hurt and is just good defensive programming.
• Initialize yaw from calibrated magnetometer data
  • Add magnetometer biases (mag_bias/x and mag_bias/y) for hard-iron compensation.
  • Initialize yaw orientation from magnetometer reading.
  • Add imu/rpy/raw and imu/rpy/filtered as debug topics. imu/rpy/raw can be used for computing magnetometer biases. imu/rpy/filtered topic is for user readability only.
• Contributors: Martin Günther, Shokoofeh Pourmehr

1.0.0 (2014-09-03)

• First public release
• Remove setting imu message frame to fixed/odom
• CMakeLists: remove unnecessary link_directories, LIBRARY_OUTPUT_PATH
• add missing build dependency on generated config This removes a racing condition from the build process.
• install nodelet xml file Otherwise the nodelet can't be found
• fix implementation of invSqrt() The old invSqrt() implementation causes the estimate to diverge under constant input. The problem was the line [long i = (long)&y;]{.title-ref}, where 64 bits are read from a 32 bit number. Thanks to \@tomas-c for spotting this and pointing out the solution.
• catkinization of imu_tools metapackage
• fix typo: zeta -> zeta_
• fix initialization of initial rotation
• gyro drift correction function added in MARG implementation
• set "zeta" as a parameter for dynamic reconfigure in the .cfg file
• add new test bag: phidgets_imu_upside_down
• add parameter publish_tf When the imu is used together with other packages, such as robot_pose_ekf, publishing the transform often interferes with those packages. This parameter allows to disable tf publishing.
• add some sample imu data
• more informative constant_dt message. Reverts to 0.0 on illegal param value
• imu_filter_madgwick manifest now correctly lists the package as GPL license.
• orientation is initialized from acceleration vector on first message received
• added dynamic reconfigure for gain parameter. Added better messages about constant_dt param at startup
• the tf published is now timestamped as the imu msg, and not as now(). Also added constant dt option for the imu+mag callback
• fix the transform publish -- from the fixed frame to the frame of the imu
• add a tf broadcaster with the orientation
• as per PaulKemppi: added option to set constant dt
• walchko: Needed to add namespace: std::isnan() and needed to add rosbuild_link_boost(imu_filter signals) to CMakeLists.txt
• added sebastian's name and link to the manifest
• renamed imu_filter to imu_filter_madgwick
• Contributors: Ivan Dryanovski, Martin Günther, Mike Purvis, Sameer Parekh, TUG-DESTOP, Francisco Vina, Michael Görner, Paul Kemppi, Tomas Cerskus, Kevin Walchko