Skip to content

Integration Walkthrough: Robin-W LiDAR#

This guide walks you through how a sensor is integrated into AutoSDV, using the Seyond Robin-W LiDAR as a concrete example. By the end, you'll understand how any sensor is integrated from hardware to software.

The Physical Sensor#

Seyond Robin-W is a solid-state LiDAR sensor: - Forward-facing (120° × 25° field of view) - Range: up to 200m - Mounted on the front dock of the vehicle - Connects via Ethernet (IP: 172.168.1.10) - Powered by 12-24V DC from vehicle power supply

When the sensor is running, it streams 3D point cloud data over the network at ~10 Hz.

The Driver Package#

The driver converts the sensor's raw data into ROS 2 messages.

Location: src/sensor_component/external/seyond_ros_driver/

This is a git submodule (external package) that: 1. Connects to Robin-W over network (172.168.1.10) 2. Receives point cloud data 3. Publishes to ROS topic: /robin_lidar/points_raw

The driver outputs PointXYZIRC format (compatible with Autoware): - x, y, z - 3D position - intensity - Reflection strength - return_type - First/last return - ring - Laser channel ID

How It's Configured#

To integrate Robin-W into the vehicle, we need to tell AutoSDV: 1. Where the sensor is physically mounted 2. How to start the driver 3. What network settings to use

Physical Position#

File: src/sensor_kit/autosdv_sensor_kit_description/config/sensor_kit_calibration.yaml

sensor_kit_base_link:
  robin_lidar_link:
    x: 0.0        # Position: at front center
    y: 0.0
    z: 0.15       # 15cm above sensor kit base
    roll: 3.14159   # 180° rotation (see explanation below)
    pitch: -1.5708  # -90° rotation
    yaw: 0.0

This defines a coordinate frame called robin_lidar_link that represents the sensor's position and orientation on the vehicle.

Why the Strange Rotation?#

Robin-W uses different coordinate axes than ROS standard:

Robin-W native: 

  X = Up
  Y = Right
  Z = Forward

ROS standard (REP-103): 

  X = Forward
  Y = Left
  Z = Up

To convert Robin-W coordinates → ROS standard: 1. Roll 180°: Flip the sensor upside down 2. Pitch -90°: Rotate so forward axis aligns

Without this transformation, the point cloud would appear upside down and rotated in RViz!

Robot Description (URDF)#

File: src/sensor_kit/autosdv_sensor_kit_description/urdf/sensor_kit.xacro

<!-- Robin-W LiDAR -->
<link name="robin_lidar_link"/>

<joint name="robin_lidar_joint" type="fixed">
  <origin
    xyz="${calibration['sensor_kit_base_link']['robin_lidar_link']['x']} ..."
    rpy="${calibration['sensor_kit_base_link']['robin_lidar_link']['roll']} ..."
  />
  <parent link="sensor_kit_base_link"/>
  <child link="robin_lidar_link"/>
</joint>

This creates the robin_lidar_link frame in the robot's TF tree using the calibration values above.

Driver Launch Configuration#

File: src/sensor_kit/autosdv_sensor_kit_launch/launch/lidar.launch.xml

<group if="$(eval &quot;'$(var lidar_model)' == 'robin-w'&quot;)">

  <!-- Launch Robin-W driver -->
  <include file="$(find-pkg-share seyond_ros_driver)/launch/robin_w.launch.py">
    <arg name="config" value="$(var sensor_model_param_path)/robin_lidar.param.yaml"/>
  </include>

  <!-- Remap topic to Autoware standard -->
  <remap from="/robin_lidar/points_raw"
         to="/sensing/lidar/robin_lidar/points_raw"/>

</group>

This says: "When user selects lidar_model:=robin-w, start the Robin-W driver and remap its topic to the standard Autoware namespace."

Driver Parameters#

File: src/param/autoware_individual_params/.../robin_lidar.param.yaml

/**:
  ros__parameters:
    ip_address: "172.168.1.10"    # Robin-W network address
    port: 2368                     # UDP port
    frame_id: "robin_lidar_link"   # TF frame name
    min_range: 0.5                 # Filter points closer than 0.5m
    max_range: 200.0               # Filter points farther than 200m

These are runtime parameters passed to the driver when it starts.

How Data Flows#

When you launch AutoSDV with lidar_model:=robin-w:

1. Robin-W Hardware
   │ Streams point cloud over UDP
   ▼
2. seyond_ros_driver Node
   │ Converts to ROS PointCloud2 message
   │ Publishes to /robin_lidar/points_raw
   ▼
3. Topic Remapping
   │ Remaps to /sensing/lidar/robin_lidar/points_raw
   ▼
4. TF Transform
   │ Applies rotation (roll=180°, pitch=-90°)
   │ Transforms points: Robin-W coords → ROS coords
   ▼
5. Autoware Localization/Perception
   │ Uses point cloud for NDT localization
   │ Detects obstacles with CenterPoint
   ▼
6. Visualization (RViz)
   │ Shows correctly oriented point cloud

How It Appears in the System#

Topic Namespace#

$ ros2 topic list | grep lidar
/sensing/lidar/robin_lidar/points_raw

All sensor data follows the pattern: /sensing/[type]/[name]/[data]

TF Tree#

$ ros2 run tf2_tools view_frames

This generates a PDF showing: 

base_link
  └─ sensor_kit_base_link
       └─ robin_lidar_link  ← Robin-W sensor frame

The transform from sensor_kit_base_link → robin_lidar_link uses the calibration values (position + rotation).

Verification#

Check that Robin-W is working:

# Topic publishing at ~10 Hz?
ros2 topic hz /sensing/lidar/robin_lidar/points_raw

# Point cloud has data?
ros2 topic echo /sensing/lidar/robin_lidar/points_raw --once

# Transform exists?
ros2 run tf2_ros tf2_echo sensor_kit_base_link robin_lidar_link

The Big Picture#

Every sensor integration follows this pattern:

  1. Physical Layer

    • Mount sensor on vehicle
    • Connect power and communication (USB/Ethernet/etc.)

  2. Driver Layer (src/sensor_component/)

    • ROS 2 package that talks to sensor hardware
    • Publishes raw sensor data to topics

  3. Sensor Kit Layer (src/sensor_kit/)

    • Calibration: Where is the sensor? (position + rotation)
    • URDF: Add sensor to robot description
    • Launch: When to start the driver
    • Parameters: Runtime configuration

  4. Data Flow

    • Driver publishes → Topics remapped → TF transforms applied → Autoware uses data

Key Takeaways#

  • Coordinate transformation matters: Robin-W's non-standard coordinates require roll=180°, pitch=-90°
  • Everything is a coordinate frame: robin_lidar_link represents the sensor in 3D space
  • Topic remapping standardizes naming: Driver outputs → /sensing/ namespace
  • Two-layer system: Drivers (sensor component) + Integration (sensor kit)

Next Steps#

Now that you understand how Robin-W is integrated: