camera_bot

Implementation of a camera-mounted steerable-wheel robot.

---

Requirements

• ROS 2 Jazzy
• ros_gz_sim
• xacro
• ros2_control, ros2_controllers, controller_manager
• A workspace with this package in src/

---

I designed the Xacro myself. I made the robot into a front-steered robot to avoid looking the same as the usual two-wheeled robot. The flaws I can see is that I'd need more time to adjust the meshes manually. Perhaps there are some ways to do it automatically.

---

1) Build

UPDATE: Docker now available

run

sudo DOCKER_BUILDKIT=1 docker build --network=host --no-cache -t camera_bot:jazzy .

After that:

# from your workspace root
source /opt/ros/$ROS_DISTRO/setup.bash
colcon build --packages-select camera_bot --cmake-clean-cache --symlink-install
. install/setup.bash

---

2) Pub/Sub

Publisher:

ros2 run camera_bot publisher \
  --ros-args -p topic:=/readings -p publish_rate_hz:=10.0

Subscriber in another terminal:

ros2 run camera_bot subscriber \
  --ros-args -p topic:=/readings -p threshold:=0.5

This pub/sub checks whether the published value exceeds the threshold.

---

3) Generate URDF for Robot State Publisher (RSP) purpose

3.1 Convert XACRO to URDF

PKG_PREFIX="$(ros2 pkg prefix camera_bot)"
XACRO_IN="${PKG_PREFIX}/share/camera_bot/description/steer.urdf.xacro"
ASSETS="file://${PKG_PREFIX}/share/camera_bot/description/assets"
PARAMS="${PKG_PREFIX}/share/camera_bot/config/controllers.yaml"

xacro "$XACRO_IN" \
  mesh_dir:="${ASSETS}" \
  ros_namespace:=ugv \
  rsp_node:=/ugv/robot_state_publisher \
  params_file:="${PARAMS}" \
  > /tmp/steer.urdf

3.2 Build /tmp/rsp.yaml from the URDF

rm -f /tmp/rsp.yaml
printf "/ugv/robot_state_publisher:\n  ros__parameters:\n    robot_description: |\n" > /tmp/rsp.yaml
tail -n +2 /tmp/steer.urdf | sed 's/^/      /' >> /tmp/rsp.yaml

---

4) Start Gazebo and spawn the robot

Gazebo:

ros2 launch ros_gz_sim gz_sim.launch.py gz_args:="-r ~/camera_bot/worlds/world.sdf"

Spawn in separate terminal:

ros2 run ros_gz_sim create -name ugv -file /tmp/steer.urdf -z 0.3

---

5) Run RSP

ros2 run robot_state_publisher robot_state_publisher \
  --ros-args -r __ns:=/ugv --params-file /tmp/rsp.yaml

---

6) Controllers

Create minimal Joint State Broadcaster (JSB) params:

cat > /tmp/jsb.yaml <<'YAML'
joint_state_broadcaster:
  ros__parameters:
    type: joint_state_broadcaster/JointStateBroadcaster
    interfaces: [position, velocity]
YAML

Spawn JSB:

ros2 run controller_manager spawner joint_state_broadcaster \
  -c /ugv/controller_manager --param-file /tmp/jsb.yaml --controller-manager-timeout 20.0

Set controller types:

ros2 param set /ugv/controller_manager steering_controller.type position_controllers/JointGroupPositionController
ros2 param set /ugv/controller_manager rear_wheels_controller.type velocity_controllers/JointGroupVelocityController

Ensure controllers are loaded:

ros2 control load_controller -c /ugv/controller_manager steering_controller    2>/dev/null || true
ros2 control load_controller -c /ugv/controller_manager rear_wheels_controller 2>/dev/null || true

Provide required params:

ros2 param set /ugv/steering_controller    joints "['steer_fl','steer_fr']"
ros2 param set /ugv/rear_wheels_controller joints "['spin_rl','spin_rr']"

Configure and activate:

ros2 control set_controller_state -c /ugv/controller_manager steering_controller inactive
ros2 control set_controller_state -c /ugv/controller_manager steering_controller active
ros2 control set_controller_state -c /ugv/controller_manager rear_wheels_controller inactive
ros2 control set_controller_state -c /ugv/controller_manager rear_wheels_controller active

Verify:

ros2 control list_controllers -c /ugv/controller_manager
ros2 control list_hardware_interfaces -c /ugv/controller_manager | sed -n '/command interfaces/,$p'

---

7) Manual command test

# steering angles (radians) for both front knuckles
ros2 topic pub -1 /ugv/steering_controller/commands std_msgs/msg/Float64MultiArray "data: [0.2, 0.2]"

# rear wheels angular velocity (rad/s)
ros2 topic pub -1 /ugv/rear_wheels_controller/commands std_msgs/msg/Float64MultiArray "data: [4.0, 4.0]"

---

Activate manual controller. This utilizes teleop.cpp

1. Source the shell

. install/setup.bash

2. Verify existence of camera_bot

ros2 pkg list | grep camera_bot
ros2 pkg prefix camera_bot

3. Run

ros2 run camera_bot teleop \
  --ros-args \
  -p steering_topic:=/ugv/steering_controller/commands \
  -p rear_topic:=/ugv/rear_wheels_controller/commands

---

Run Camera

First, run the bridge for clock to enable use_sim_time

ros2 run ros_gz_bridge parameter_bridge /clock@rosgraph_msgs/msg/Clock@gz.msgs.Clock

Then run the bridge for camera

ros2 run ros_gz_bridge parameter_bridge \
  /ugv/camera/image@sensor_msgs/msg/Image@gz.msgs.Image \
  /ugv/camera/depth_image@sensor_msgs/msg/Image@gz.msgs.Image \
  /ugv/camera/camera_info@sensor_msgs/msg/CameraInfo@gz.msgs.CameraInfo \
  /clock@rosgraph_msgs/msg/Clock@gz.msgs.Clock

These commands run the Visual Odometry and SLAM

In two separate terminals, run tf2 publisher:

ros2 run tf2_ros static_transform_publisher 0 0 0 0 0 0 body ugv/body/rgb_camera

ros2 run tf2_ros static_transform_publisher 0 0 0 0 0 0 body ugv/body/depth_camera

run rgbd_odometry for visual odometry:

ros2 run rtabmap_odom rgbd_odometry --ros-args \
  -p use_sim_time:=true \
  -p frame_id:=body \
  -p approx_sync:=true \
  -p qos_image:=2 -p qos_camera_info:=2 \
  -p wait_for_transform:=0.2 \
  -r rgb/image:=/ugv/camera/image \
  -r depth/image:=/ugv/camera/depth_image \
  -r rgb/camera_info:=/ugv/camera/camera_info

For the RTAB-Map, parse this YAML file first to avoid data type mismatch:

cat > /tmp/rtabmap.yaml <<'YAML'
/rtabmap:
  ros__parameters:
    use_sim_time: true
    frame_id: "body"
    odom_frame_id: "odom"
    map_frame_id: "map"
    subscribe_depth: true
    approx_sync: true
    qos_image: 2
    qos_camera_info: 2
    RGBD/CreateOccupancyGrid: "true"
    Grid/FromDepth: "true"
    Grid/RangeMax: "8.0"
    Grid/CellSize: "0.05"
YAML

then run

ros2 run rtabmap_slam rtabmap --ros-args \
  --params-file /tmp/rtabmap_3d.yaml \
  -r rgb/image:=/ugv/camera/image \
  -r depth/image:=/ugv/camera/depth_image \
  -r rgb/camera_info:=/ugv/camera/camera_info

All of these can now be checked in Rviz2, check PointCloud2 to see SLAM capability. There are still some issues with the SLAM that I need to ensure.

Troubleshooting

• Package not visible:

  source /opt/ros/$ROS_DISTRO/setup.bash
  . install/setup.bash
  ros2 pkg list | grep camera_bot

• Cache/source mismatch errors: clean the workspace root

  rm -rf build/ install/ log/
  colcon build --packages-select camera_bot --symlink-install

• No subscribers on commands: ensure /ugv namespace is used and controllers are active.

To-be-fixed

• Fix SLAM capability.
• Fix the mesh so the steering bot will look a bit more proper than a slab with wheels.