NVblox fails to load map in dynamic mode

Robotics & Edge Computing Robotics - Isaac Isaac ROS
1

I’m using NVblox in ROS Humble for robot navigation. When setting it to dynamic mode, there are several issues with saving and loading the map:

  1. Mesh and freespace voxel layers are not serialized and saved by the save_map service. Calling the service produces the errors listed below (in addition to the normal output):

    [nvblox_node-1] E0901 12:54:48.250876 550 serializer.cpp:126] Unrecognized layer type, can’t serialize: N6nvblox10BlockLayerINS_9MeshBlockEEE
    [nvblox_node-1] E0901 12:54:48.314859 550 serializer.cpp:126] Unrecognized layer type, can’t serialize: N6nvblox15VoxelBlockLayerINS_14FreespaceVoxelEEE

  2. When trying to load the map (using the load_map service), it tries to find the freespace voxel layer, which was not saved (see #1), and fails, killing the NVblox node. Here is the load_map output:

    [nvblox_node-1] W0908 16:19:58.914960 552 layer_cake_impl.h:86] Request for a LayerType: N6nvblox15VoxelBlockLayerINS_14FreespaceVoxelEEE which is not in the cake.
    [nvblox_node-1] F0908 16:19:58.916170 552 layer_cake_impl.h:110] Check failed: ‘ptr’ Must be non NULL
    [nvblox_node-1] *** Check failure stack trace: ***
    [nvblox_node-1] @ 0x7ed0b516cb03 google::LogMessage::Fail()
    [nvblox_node-1] @ 0x7ed0b51749d1 google::LogMessage::SendToLog()
    [nvblox_node-1] @ 0x7ed0b516c7c2 google::LogMessage::Flush()
    [nvblox_node-1] @ 0x7ed0b516e78f google::LogMessageFatal::~LogMessageFatal()
    [nvblox_node-1] @ 0x7ed09f7324b6 nvblox::LayerCake::get<>()
    [nvblox_node-1] @ 0x7ed09f722322 nvblox::Mapper::updateEsdfSlice()
    [nvblox_node-1] @ 0x7ed09f74b3eb nvblox::MultiMapper::updateEsdfOfMapper()
    [nvblox_node-1] @ 0x7ed09f74b970 nvblox::MultiMapper::updateEsdf()
    [nvblox_node-1] @ 0x7ed0b2cbc3c4 nvblox::NvbloxNode::processEsdf()
    [nvblox_node-1] @ 0x7ed0b2cafa47 nvblox::NvbloxNode::tick()
    [nvblox_node-1] @ 0x7ed0b2cc1176 rclcpp::GenericTimer<>::execute_callback()
    [nvblox_node-1] @ 0x7ed09f4f0301 rclcpp::Executor::execute_any_executable()
    [nvblox_node-1] @ 0x7ed09f4f76ea rclcpp::executors::MultiThreadedExecutor::run()
    [nvblox_node-1] @ 0x7ed09eeb0253 (unknown)
    [nvblox_node-1] @ 0x7ed09ec3fac3 (unknown)
    [nvblox_node-1] @ 0x7ed09ecd1850 (unknown)
    dev@STAS-NOTEBOOK:/workspace$ [ERROR] [nvblox_node-1]: process has died [pid 65, exit code -6, cmd ‘/workspace/install/nvblox_ros/lib/nvblox_ros/nvblox_node --ros-args -r __node:=nvblox_node -r __ns:=/spectra --params-file /workspace/install/aivot_robot_spectra/share/aivot_robot_spectra/config/nvblox_base.yaml --params-file /tmp/launch_params_qjtd3s4c --params-file /tmp/launch_params_erep3gss --params-file /tmp/launch_params_efwcxr8c -r /tf:=tf -r /tf_static:=tf_static -r camera_0/depth/image:=camnav/depth/depth_registered -r camera_0/depth/camera_info:=camnav/zed/depth/camera_info -r camera_0/color/image:=camnav/left/image_rect_color -r camera_0/color/camera_info:=camnav/left/camera_info -r /tf:=tf -r /tf_static:=tf_static’].

How can I fix this?

3

Hi @Robotmaker2005,

Thank you for your post.

This issue should have been fixed in February as part of a minor update for Isaac ROS 3.2. Are you using an earlier version? Which demo are you using to reproduce the error?

Best,
Ahung

5

My NVblox version is 3.2.5.
I’m using a custom setup, not a demo. Here’s my NVblox config (contents of the nvblox_base.yaml):

nvblox_base.txt (4.5 KB)

6

Hi @Robotmaker2005

Thank you for the info.
I can reproduce the error at my end. Let me forward this issue to our engineers.

Best,
Ahung

1 Like
9

Hi @Robotmaker2005,

I’d like to update you that the fix has been released. Please pull the latest 3.2 release of isaac_ros_nvblox and give it a try.

Best,
Ahung

1 Like
10

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.