Sensor Simulation Skill

Persona

Think like a robotics perception engineer who integrates sensors into robot systems. You understand sensor physics, noise models, and data visualization. You configure sensors that produce realistic data streams suitable for perception algorithm development.

Pre-Flight Questions

Before adding or configuring any sensor, ask yourself:

1. Sensor Purpose

Q: What does the robot need to perceive?
- Visual: Camera for images, object detection
- Distance: LIDAR for mapping, obstacle detection
- Motion: IMU for orientation, acceleration
- Touch: Contact for collision detection

2. Sensor Placement

Q: Where should the sensor be mounted?
- Impact: FOV coverage, occlusion, stability
- Forward-facing camera → front of robot
- 360° LIDAR → top center, unobstructed
- IMU → center of mass, rigidly mounted

3. Performance Trade-offs

Q: What resolution/rate is needed?
- High resolution: Better data, more computation
- High rate: Better responsiveness, more bandwidth
- Balance based on application requirements

Principles

Principle 1: Sensor Structure in SDF

Sensors are added to links within robot models:

<link name="camera_link">
  <!-- Visual and collision for the sensor housing -->
  <visual>...</visual>
  <collision>...</collision>
  <inertial>...</inertial>

  <!-- The sensor itself -->
  <sensor name="camera" type="camera">
    <camera>
      <horizontal_fov>1.047</horizontal_fov>
      <image>
        <width>640</width>
        <height>480</height>
      </image>
      <clip>
        <near>0.1</near>
        <far>100</far>
      </clip>
    </camera>
    <always_on>1</always_on>
    <update_rate>30</update_rate>
    <topic>camera/image</topic>
  </sensor>
</link>

Principle 2: Camera Configuration

<sensor name="front_camera" type="camera">
  <camera>
    <!-- Field of view in radians (60° = 1.047 rad) -->
    <horizontal_fov>1.047</horizontal_fov>

    <!-- Image dimensions -->
    <image>
      <width>640</width>
      <height>480</height>
      <format>R8G8B8</format>
    </image>

    <!-- Depth range -->
    <clip>
      <near>0.1</near>
      <far>100</far>
    </clip>
  </camera>

  <always_on>1</always_on>
  <update_rate>30</update_rate>
  <visualize>true</visualize>
  <topic>camera/image</topic>
</sensor>

Common resolutions:

640x480: Standard, good performance
1280x720: HD, more detail
320x240: Low-res, fast processing

Principle 3: LIDAR Configuration

<sensor name="lidar" type="gpu_lidar">
  <lidar>
    <scan>
      <horizontal>
        <samples>360</samples>
        <resolution>1</resolution>
        <min_angle>-3.14159</min_angle>
        <max_angle>3.14159</max_angle>
      </horizontal>
      <vertical>
        <samples>1</samples>
        <resolution>1</resolution>
        <min_angle>0</min_angle>
        <max_angle>0</max_angle>
      </vertical>
    </scan>
    <range>
      <min>0.1</min>
      <max>10</max>
      <resolution>0.01</resolution>
    </range>
    <noise>
      <type>gaussian</type>
      <mean>0</mean>
      <stddev>0.01</stddev>
    </noise>
  </lidar>

  <always_on>1</always_on>
  <update_rate>10</update_rate>
  <visualize>true</visualize>
  <topic>lidar/scan</topic>
</sensor>

Key parameters:

samples: Points per scan (more = denser)
min/max_angle: Scan coverage (-π to π = 360°)
min/max range: Detection distance limits
noise: Gaussian noise for realism

Principle 4: IMU Configuration

<sensor name="imu" type="imu">
  <imu>
    <angular_velocity>
      <x>
        <noise type="gaussian">
          <mean>0</mean>
          <stddev>0.01</stddev>
        </noise>
      </x>
      <!-- y and z similar -->
    </angular_velocity>
    <linear_acceleration>
      <x>
        <noise type="gaussian">
          <mean>0</mean>
          <stddev>0.1</stddev>
        </noise>
      </x>
      <!-- y and z similar -->
    </linear_acceleration>
  </imu>

  <always_on>1</always_on>
  <update_rate>100</update_rate>
  <topic>imu/data</topic>
</sensor>

IMU outputs:

angular_velocity: Rotation rate (rad/s)
linear_acceleration: Includes gravity!
orientation: Quaternion (if available)

Principle 5: Add Noise for Realism

Real sensors have noise. Simulated sensors should too:

<noise>
  <type>gaussian</type>
  <mean>0</mean>
  <stddev>0.01</stddev>  <!-- Adjust based on sensor quality -->
</noise>

Typical noise levels:

High-quality LIDAR: stddev 0.005-0.01m
Consumer LIDAR: stddev 0.02-0.05m
IMU gyro: stddev 0.01-0.05 rad/s
IMU accel: stddev 0.1-0.5 m/s²

Debugging Sensor Issues

Sensor Not Publishing

Check topic name matches subscriber
Verify <always_on>1</always_on>
Confirm sensor plugin is loaded
Check gz topic -l for available topics

Data All Zeros

Sensor origin inside robot body (collision)
Nothing in sensor's field of view
Range limits too restrictive

Noisy/Unstable Data

Noise parameters too high
Physics timestep too large
Sensor update rate too fast

Checklist

Before finalizing any sensor configuration:

[ ] Sensor attached to correct link
[ ] Origin places sensor at correct position
[ ] FOV/range covers intended area
[ ] Update rate appropriate for application
[ ] Noise model added for realism
[ ] Topic name is unique and descriptive
[ ] visualize enabled for debugging
[ ] Data appears in gz topic -e
[ ] Data format matches ROS 2 expectations

Integration

This skill is used by:

content-implementer agent when generating Module 2 lessons
Students learning sensor simulation in Chapter 11
Perception algorithm development and testing

Dependencies:

urdf-robot-model - sensors attach to robot links
gazebo-world-builder - sensors perceive the world
ros2-gazebo-bridge - sensor data bridges to ROS 2