i/rtabmap
1
0
Fork 0
mirror of https://github.com/introlab/rtabmap.git synced 2026-10-12 21:09:50 +08:00
Code Issues Packages Projects Releases Wiki Activity

Imu motion predictor (#1791)

Browse Source 
* Imu motion predictor

* simplified parameters

* refactored odom's imu init orientation

* fixing sign issue with x86-64-v3

* fixing M_PI not defined in tests

* Added auto gravity estimation mode

* Fixing crash on windows when uFormat() is to long
This commit is contained in:
matlabbe
2026-10-11 13:59:27 -07:00
committed by GitHub
parent 8d93f807c8
commit dca24fc11f
20 changed files with 1777 additions and 132 deletions
Download Patch File Download Diff File Expand all files Collapse all files
guilib/src/PreferencesDialog.cpp
+1
View File
@@ -1504,6 +1504,7 @@ PreferencesDialog::PreferencesDialog(QWidget * parent) :
_ui->odom_flow_scanKeyframeThr->setObjectName(Parameters::kOdomScanKeyFrameThr().c_str());
_ui->odom_flow_guessMotion->setObjectName(Parameters::kOdomGuessMotion().c_str());
_ui->odom_guess_smoothing_delay->setObjectName(Parameters::kOdomGuessSmoothingDelay().c_str());
_ui->odom_imu_gravity->setObjectName(Parameters::kOdomImuGravity().c_str());
_ui->odom_imageDecimation->setObjectName(Parameters::kOdomImageDecimation().c_str());
_ui->odom_alignWithGround->setObjectName(Parameters::kOdomAlignWithGround().c_str());
_ui->odom_lidar_deskewing->setObjectName(Parameters::kOdomDeskewing().c_str());
guilib/src/ui/preferencesDialog.ui
+50 -15
View File
@@ -16427,7 +16427,7 @@ With <0, the length is estimated once for each unique marker, then re-used fo
</property>
</widget>
</item>
<item row="14" column="1">
<item row="15" column="1">
<widget class="QLabel" name="label_232">
<property name="text">
<string>Data buffer size (0 means inf).</string>
@@ -16570,7 +16570,7 @@ With &lt;0, the length is estimated once for each unique marker, then re-used fo
</property>
</widget>
</item>
<item row="12" column="0">
<item row="13" column="0">
<widget class="QDoubleSpinBox" name="odom_flow_keyframeThr">
<property name="maximum">
<double>1.000000000000000</double>
@@ -16590,7 +16590,7 @@ With &lt;0, the length is estimated once for each unique marker, then re-used fo
</property>
</widget>
</item>
<item row="11" column="0">
<item row="12" column="0">
<widget class="QSpinBox" name="odom_VisKeyFrameThr">
<property name="maximum">
<number>9999</number>
@@ -16600,7 +16600,7 @@ With &lt;0, the length is estimated once for each unique marker, then re-used fo
</property>
</widget>
</item>
<item row="13" column="1">
<item row="14" column="1">
<widget class="QLabel" name="label_246">
<property name="text">
<string>[Geometry] Create a new keyframe when the number of inliers drops under this threshold. Setting value to 0 means that a keyframe is created for each processed frame.</string>
@@ -16613,7 +16613,7 @@ With &lt;0, the length is estimated once for each unique marker, then re-used fo
</property>
</widget>
</item>
<item row="10" column="1">
<item row="11" column="1">
<widget class="QLabel" name="label_248">
<property name="text">
<string>Decimation of the RGB image before registration. If depth size is larger than decimated RGB size, depth is decimated to be always at most equal to RGB size. If Visual Registration -&gt; Visual Feature -&gt; Depth as Mask is true and if depth is smaller than decimated RGB, depth may be interpolated to match RGB size for feature detection.</string>
@@ -16709,7 +16709,7 @@ With &lt;0, the length is estimated once for each unique marker, then re-used fo
</property>
</widget>
</item>
<item row="15" column="0">
<item row="16" column="0">
<widget class="QPushButton" name="pushButton_testOdometry">
<property name="text">
<string>Test odometry</string>
@@ -16729,14 +16729,14 @@ With &lt;0, the length is estimated once for each unique marker, then re-used fo
</property>
</widget>
</item>
<item row="14" column="0">
<item row="15" column="0">
<widget class="QSpinBox" name="odom_dataBufferSize">
<property name="maximum">
<number>999999</number>
</property>
</widget>
</item>
<item row="12" column="1">
<item row="13" column="1">
<widget class="QLabel" name="label_196">
<property name="text">
<string>[Visual] Create a new keyframe when the number of inliers drops under this ratio of features in last frame. Setting the value to 0 means that a keyframe is created for each processed frame.</string>
@@ -16749,7 +16749,7 @@ With &lt;0, the length is estimated once for each unique marker, then re-used fo
</property>
</widget>
</item>
<item row="10" column="0">
<item row="11" column="0">
<widget class="QSpinBox" name="odom_imageDecimation">
<property name="minimum">
<number>1</number>
@@ -16778,7 +16778,7 @@ With &lt;0, the length is estimated once for each unique marker, then re-used fo
</property>
</widget>
</item>
<item row="13" column="0">
<item row="14" column="0">
<widget class="QDoubleSpinBox" name="odom_flow_scanKeyframeThr">
<property name="maximum">
<double>1.000000000000000</double>
@@ -16794,7 +16794,7 @@ With &lt;0, the length is estimated once for each unique marker, then re-used fo
<item row="8" column="1">
<widget class="QLabel" name="label_520">
<property name="text">
<string>Guess smoothing delay (s). Estimated velocity is averaged based on last transforms up to this maximum delay. This can help to get smoother velocity prediction. Last velocity computed is used directly if a filtering strategy is set or the delay is below the odometry rate.</string>
<string>Guess smoothing delay (s). The velocity is averaged over the last transforms up to this delay, for a smoother velocity prediction. The last velocity is used directly if a filtering strategy is set or the delay is below the odometry rate. With an IMU giving orientation and linear acceleration, a delay &gt; 0 also enables the IMU acceleration: the velocity is the displacement over this delay corrected by the acceleration measured since, so it is not delayed, and the motion guess and lidar deskewing integrate the acceleration from it (~0.5 s recommended). With 0, the IMU only gives the orientation. Without IMU, the average is delayed by half this delay: useful for platforms with inertia, high frame rates, or lidar deskewing, where the velocity noise would feed back into the next poses.</string>
</property>
<property name="wordWrap">
<bool>true</bool>
@@ -16831,7 +16831,7 @@ With &lt;0, the length is estimated once for each unique marker, then re-used fo
</property>
</widget>
</item>
<item row="11" column="1">
<item row="12" column="1">
<widget class="QLabel" name="label_354">
<property name="text">
<string>[Visual] Create a new keyframe when the number of inliers drops under this threshold. Setting the value to 0 means that a keyframe is created for each processed frame.</string>
@@ -16844,10 +16844,10 @@ With &lt;0, the length is estimated once for each unique marker, then re-used fo
</property>
</widget>
</item>
<item row="9" column="1">
<item row="10" column="1">
<widget class="QLabel" name="label_7461">
<property name="text">
<string>Lidar deskewing. If input lidar has time channel, it will be deskewed with a constant motion model (with IMU orientation and/or guess if provided).</string>
<string>Lidar deskewing. If input lidar has time channel, it will be deskewed. With an IMU, the pose of every point is predicted from the previous frame: orientation from the IMU, translation from the velocity (with the IMU acceleration if the guess smoothing delay is &gt; 0). Without IMU, with a constant motion model (or the guess if provided).</string>
</property>
<property name="wordWrap">
<bool>true</bool>
@@ -16857,13 +16857,48 @@ With &lt;0, the length is estimated once for each unique marker, then re-used fo
</property>
</widget>
</item>
<item row="9" column="0">
<item row="10" column="0">
<widget class="QCheckBox" name="odom_lidar_deskewing">
<property name="text">
<string/>
</property>
</widget>
</item>
<item row="9" column="0">
<widget class="QDoubleSpinBox" name="odom_imu_gravity">
<property name="specialValueText">
<string>Auto</string>
</property>
<property name="suffix">
<string> m/s²</string>
</property>
<property name="decimals">
<number>5</number>
</property>
<property name="maximum">
<double>100.000000000000000</double>
</property>
<property name="singleStep">
<double>0.010000000000000</double>
</property>
<property name="value">
<double>9.806650000000000</double>
</property>
</widget>
</item>
<item row="9" column="1">
<widget class="QLabel" name="label_odom_imu_gravity">
<property name="text">
<string>Gravity magnitude removed from the IMU linear acceleration (used with a guess smoothing delay &gt; 0 above). Standard gravity by default. Set it to what the accelerometer reads at rest to compensate for its scale error, or for another gravity than Earth's. Auto (0): estimated from the IMU while it is still, with standard gravity until then: the robot should then be perfectly still for a moment (e.g., at start).</string>
</property>
<property name="wordWrap">
<bool>true</bool>
</property>
<property name="textInteractionFlags">
<set>Qt::LinksAccessibleByMouse|Qt::TextSelectableByMouse</set>
</property>
</widget>
</item>
</layout>
</item>
<item>