Accurate Docking for Mobile Robots with Triton

Picking up goods often requires a high level of repeat accuracy. Technologies like LiDAR-localization can struggle with accuracy in dynamic environments. Other technologies depend on infrastructure such as magnetic tape and codes (which is expensive to commission and difficult to maintain). Hence in this article we explain how Triton Floor-LOC can be used for docking with high accuracy.

Figure: Kumatech Tractor Tom vehicle using teach and repeat docking to pick up goods.

1. Vehicle sends external reference to Triton (LiDAR-based localization, wheel odom, ..)
2. 1. Info: For this guide we will be using SICK’s LiDAR-LOC 2.10 for LiDAR-localization. This automatically sends the external reference to Triton.
   2. Tip: if your LiDAR-localization is noisy, consider driving the vehicle to the start position, then disable SLAM updates and rely on wheel odometry (smooth) to map the line, while having some sense of global accuracy
3. Position vehicle in end position of docking stations.
4. 1. It is important to position the vehicle as accurate as possible in the end position of the docking station.
5. Write down the robot pose belonging to this docking position.

Figure: Robot is placed in the docking station. The pose of the robot is marked in red and will later be used to draw a navigation path.

4. Enable mapping mode on some unused cluster ID
5. Drive vehicle slowly backwards in a straight line out of the docking station. Best practice to have line length of at least 3 times the vehicle length
   ​Note: When driving backwards out of the docking station, try to drive as straight as possible.

Figure: Robot is driven out of the docking station.The pose of the robot is marked in red and will later be used to draw a navigation path.

6. Disable mapping mode
7. Write down the exact robot pose belonging to this position.
8. Optional: Mark end position/contour of vehicle using tape (easier to localize vehicle later on while testing, or check if vehicle is on the line or not)

Video: Vehicle starts from docking stations and drive backwards while Triton is mapping the floor. LiDAR-localization from LiDAR-LOC is used as external reference. Mapping mode is enabled via SOPAS-Air.

• Ensure Triton corrections are used to update the robot pose and external reference

Note: LiDAR-LOC 2.10 will automatically include Triton corrections in its localization and send the external reference to Triton.

• For the docking use-case our goal is to drive vehicle into and out of docking station.
• In order to do so, you need to define the start pose and end pose of our docking
• In general we recommend to use the start and end pose you wrote down in procedure 2.1. Mapping.

Video: A navigation path is added based on the begin and end pose from the teach & repeat docking procedure. This docking path is then connected to the main navigation paths.

1. Determine a path based on the start and end pose you found in step 2.3.1.
2. 1. If you drove perfectly straight and used LiDAR-localization for mapping, you can draw a straight line.
   2. If you mapped a cluster with a curve or used wheel/Triton odometry for mapping (mapped cluster looks like a curve due to odometry drift), you might need to define a more complex path (curve, polynomial, …).
3. Navigate along the line using your navigation software.
4. Stop the vehicle when it reaches the target position your wrote down in procedure 2.1. Mapping

Note: For LiDAR-LOC users we offer a Virtual Line Navigation (VLN) extension. This extension allows you to configure virtual line sensors that will provide data to let your mobile robot follow the navigation paths. Click here to learn more about the VLN extension.

Video: Vehicle drives along navigation path to the end position in the docking station. A Triton map is present for the last meter to ensure accurate localization.

• In the procedure above you learned how to dock with high accuracy using Triton.
• In other parts of your factory/warehouse you can also use Triton to achieve reliable localization, but you might choose to navigate based on LiDAR-localization only.
• In that case however, switching from LiDAR-localization to Triton for docking poses the following challenges:
• • Triton can match if Triton is physically within 3cm of a mapped line (due to Triton’s limited field-of-view)
  • You need to update your robot’s pose and external reference based on Triton poses
  • • Note: LiDAR-LOC 2.10 will automatically include Triton corrections in its localization and send the external reference to Triton.

If you observed that the localization and/or navigation of your robot is not accurate enough to drive to the Triton line and get Triton poses, you can map extra clusters to assist with the transition:

Video: You can map extra help clusters in case your robot's localization and/or navigation is not accurate enough to 'find' the Triton line. Make sure you are on top of the Triton line, get poses from Triton, update your robot's localization and external reference to Triton before starting to map a help clusters. LiDAR-LOC 2.10 automatically updates the robot pose and external reference based on Triton poses.