NVIDIA-ISAAC-ROS / isaac_ros_visual_slam

Visual SLAM/odometry package based on NVIDIA-accelerated cuVSLAM
https://developer.nvidia.com/isaac-ros-gems
Apache License 2.0
904 stars 142 forks source link

IMU fusion status #148

Open dakejahl opened 9 months ago

dakejahl commented 9 months ago

Is there any way to tell if IMU fusion is working or not? The VisualSlamStatus.vo_state doesn't ever change from 1.

# Pure visual odometry return code:
# 0 - Unknown state
# 1 - Success
# 2 - Failed
# 3 - Success but invalidated by IMU
uint8 vo_state

Same goes for when tracking is lost, vo_state remains at 1.

AlexKlimaj commented 9 months ago

We are testing on a drone, using the output of VIO to hold position with PX4. Knowing the state of would be very helpful in PX4.

https://github.com/ARK-Electronics/isaac_ros-dev

https://youtube.com/playlist?list=PLUepQApgwSox1ydUYJx27fTYQnn74SkTY&si=Yk6WCbx0eF9s6Y97

https://review.px4.io/plot_app?log=38789a73-c5d6-4702-bbea-1f87ce57f67a

hemalshahNV commented 9 months ago

IMU fusion is actually a feature we will ship in the next release of Isaac ROS. The interface had been added for compatibility with other editions of the libcuvslam libraries.

AlexKlimaj commented 9 months ago

So this statement in the README isn't correct about using IMU?

This package uses a stereo camera with an IMU to estimate odometry as an input to navigation.

hemalshahNV commented 9 months ago

There was an IMU fallback feature (when visual tracking is lost, libcuvslam would try to run on pure IMU for a fix timebound), but that's been replaced with IMU fusion.

hermanndererdmann commented 8 months ago

is there a timeframe when the new release will be published?

franfrancisco9 commented 5 months ago

Any news on when IMU Fusion will be active?

anthonywebb commented 3 months ago

It looks like Isaac ROS 3 added it, but I can confirm it still doesnt work

yurii-drevnytskyi-dianomix commented 1 month ago

Can you confirm that IMU fusion doesn't work? I'm trying to figure out why turning it on doesn't improve the odometry results.