Help with TF Lookup Error in NvbloxCostmapLayer: "Extrapolation into the Future" Issue

Robotics & Edge Computing Robotics - Isaac Isaac ROS
1

Dear Isaac ROS Team @Raffaello

I am encountering an issue with the NvbloxCostmapLayer when trying to transform between the map and odom frames in my ROS2-based application. The system frequently logs a warning about extrapolation into the future, as shown below:
[component_container_isolated-4] [WARN] [1727793921.601548454] [global_costmap.global_costmap]: [NvbloxCostmapLayer] Can’t transform: map to odom. Error: Lookup would require extrapolation into the future. Requested time 1727793921.600585 but the latest data is at time 1727793921.580367, when looking up transform from frame [odom] to frame [map]

It seems that the requested time for the transform is slightly ahead of the latest available data, causing the TF lookup to fail.

For context, I am using nvblox and have enabled visual odometry. I am successfully publishing the transform relationships from map to odom and odom to base_link. Despite this, the issue occurs intermittently.

Could you provide any guidance or best practices to avoid this “extrapolation into the future” error in Isaac ROS with NvbloxCostmapLayer? Specifically, I would like to know:

  • Are there recommended TF buffer or transform settings that can help mitigate this issue?
  • Should I be handling the transform requests differently to account for time delays or synchronization issues?
  • Any suggestions on improving TF performance, especially with visual odometry, to prevent this issue?

image
image1920×821 105 KB

image
image1920×144 41.6 KB

Thank you in advance for your help!

1 Like
2

Hi @alex.jaywill

Thank you for your post.

These types of errors are usually related to communication. There is one or more nodes that are receiving TF in delay. It can be the hardware communication, compute time, or something around these points.

I have some questions to help you better.

Let me know.

Raffaello

3

Hi @Raffaello. I am having the same issue as well. All nodes are running on the same platform, and I’m using a Jetson Orin Nano. I tried the two solutions in the linked page, but neither worked.

4

Hi @awi2

To better help you, have you tried the same setup running everything on the same machine?

Does this error also appear when you run our demo?

Best,
Raffaello

5

I am currently running everything on the same machine, and the issue appears with the demo as well.

The error happens with a rosbag and live on the real sense.

6

@Raffaello I noticed the global_costmap is publishing at a lower rate than the Nav2 parameter, but changing the parameter doesn’t help. Wondering where else the problem may be.

7

The issue may be related to the Jetson Orin Nano, which doesn’t have enough power to publish all packages simultaneously.

Try to switch off other software that is not required on your device and/or also switch off the User Interface.

Best,
Raffaello

8

Hi , Iam facing the same issue pretty much , were you able to solve it?

9

Ho @flavionzambi

Looking your errors related to this topic: Nvblox Nav2 Plugin Implementation trouble

seems your NVIDIA Jetson is overloaded and doesn’t have enough time to process all data.

Best,
Raffaello

1 Like