SLAM for Agriculture vehicle with Dual antenna RTK GPS system.

Software ROS2
,
1

Dear ZED Community,

We are currently working with a vehicle(ROS2) equipped with a ZED2i camera and a dual-antenna RTK GPS system (for precise positional accuracy and robust heading estimation). Our present localization pipeline relies entirely on GPS, which works well in open-sky outdoor environments but prevents us from operating reliably in areas with limited satellite visibility.

We are now exploring a hybrid indoor–outdoor setup. While reviewing potential solutions, we noticed that ZED provides its own SLAM-based positional estimates directly through the driver.

I have a few questions regarding this:

  1. GPS + Heading Input:
    The ZED driver currently accepts only /gps/fix, but our RTK system also outputs highly accurate heading information.
    Is there a recommended way to fuse or integrate external heading data with ZED’s tracking to improve overall consistency, especially during outdoor operation?

  2. Industry Use Cases:
    Are there any known industrial or agricultural platforms actively using ZED SLAM for field-level navigation or mixed indoor–outdoor applications?
    Real-world references would be extremely helpful.

  3. Geotracking Resources:
    I’m looking for detailed guidance or examples on performing geo-referenced tracking using the ZED camera—particularly best practices for aligning ZED SLAM output with RTK world coordinates.
    If there are any tutorials, documentation, or community projects covering this, please share them.

Thank you for your time and support.

Regards,
Raman

2

Hi @RamanaBotta,

Welcome to the forums :waving_hand:

To answer your questions:

  1. Even though the heading cannot be supplied directly to the ZED node, there is still a practical way to benefit from it. The typical approach is to let the ZED perform its own GNSS fusion to produce globally consistent visual–inertial poses, and then feed those fused poses, along with your RTK heading, into an external state estimator, usually an EKF such as robot_localization. In this configuration, the ZED provides smooth, drift-free global position updates, while the RTK system’s heading acts as a highly accurate orientation reference. When combined, they reinforce each other and deliver a more stable global estimate than either source alone.
  2. I cannot necessarily cite many of our customers, but you can find this case study from IndroRobotics that share how they use the ZED SDK’s SLAM with their own localization stack
  3. You can find our documentation on GNSS fusion with ZED SDK SLAM here: Global Localization Overview - Stereolabs. You can find more tutorials and code samples in this directory of our Github: zed-sdk/global localization at master · stereolabs/zed-sdk · GitHub
1 Like