roscon2025_rbwatcher_workshop

RB-Watcher ROS 2 digital twin for autonomous inspection at an electrical substation. The project combines Gazebo Harmonic simulation, the Nav2 navigation stack, and BehaviorTree.CPP to demonstrate guided patrol and reactive PTZ tracking workflows.

---

Table of Contents

1. Overview
2. Workshop Objective
3. Requirements
4. Workspace Setup
5. Build Instructions
6. Repository Structure
7. Behavior Tree Tooling
8. Workshop Tasks
   • Task 1 – Launch the Simulation
   • Task 2 – Controllers and Sensors
   • Task 3 – Basic Control
   • Task 4 – Mapping
   • Task 5 – Localization & Navigation
   • Task 6 – Perception
   • Task 7 – PTZ Person Tracking
   • Task 8 – Autonomous Inspection with Behavior Trees
9. Open Exercises

---

Overview

The RB-Watcher is a mobile platform designed for persistent inspection and surveillance missions. This workshop package provides:

• A Gazebo Harmonic simulation of the RB-Watcher operating in an electrical substation.
• Launch and configuration assets to exercise Nav2, perception, and PTZ tracking components.
• A BehaviorTree.CPP action server (rbwatcher_behaviors) showcasing patrol-and-track missions using custom tree nodes.

---

Workshop Objective

The guided exercises build toward a complete autonomous inspection workflow. Participants first stand up simulation, then generate an actionable map, localize and navigate with Nav2, plug in perception, and finally orchestrate everything with BehaviorTree-driven mission logic. By Task 8 you will understand how routing, detection, and PTZ tracking interlock inside a maintainable, modular autonomy stack.

---

Requirements

• Ubuntu 24.04 with ROS 2 Jazzy and Gazebo Harmonic pre-installed.
• Desktop-class GPU recommended for running the Gazebo world and RViz simultaneously.
• Basic familiarity with ROS 2 CLI tools and the colcon build system.

---

Workspace Setup

mkdir -p ~/workspaces/roscon_ws/src
cd ~/workspaces/roscon_ws/src
git clone --recurse-submodules -b jazzy-devel https://github.com/RobotnikAutomation/roscon2025_rbwatcher_workshop.git

Install system dependencies and resolve ROS package requirements:

cd ~/workspaces/roscon_ws
sudo apt-get update
rosdep update
rosdep install --from-paths src --ignore-src -r -y
sudo apt install -y $(find -name '*ros-jazzy-robotnik*.deb')

---

Build Instructions

cd ~/workspaces/roscon_ws
colcon build --symlink-install
source install/setup.bash

Re-source install/setup.bash in new terminals or add it to your shell profile.

---

Repository Structure

Key packages in this workspace:

• robotnik_gazebo_ignition: Robotnik's Gazebo Harmonic simulation.
• electrical_substation_world: Gazebo Harmonic environment used throughout the workshop.
• rbwatcher_description: URDF/SDF assets, meshes, and robot configuration for RB-Watcher.
• rbwatcher_behaviors: BehaviorTree.CPP action server plus custom Nav2 and PTZ tracker BT nodes.
• simple_person_detector: Simulated perception pipeline that publishes person detections.
• ptz_tracker: PTZ camera controller node that tracks detected persons using a joint trajectory controller.

Refer to each package’s README for deeper details or configuration options.

---

Behavior Tree Tooling

Visualize and edit trees with Groot2. The snippet below downloads the latest AppImage directly from GitHub releases:

cd ~/Downloads
wget <Groot2 AppImage URL>
chmod +x Groot2-*.AppImage
./Groot2-*.AppImage

---

Workshop Tasks

Task 1 – Launch the Simulation

Launch the electrical substation world:

ros2 launch robotnik_gazebo_ignition spawn_world.launch.py \
  world_path:=$(ros2 pkg prefix electrical_substation_world)/share/electrical_substation_world/worlds/electrical_substation.world \
  gui:=true

Tips:

• Set gui:=false for a headless Gazebo session.
• Override world_path to experiment with alternative scenes.
• Reduce graphics load on low-powered machines:

export LOW_PERFORMANCE_SIMULATION=true

Spawn the RB-Watcher robot:

ros2 launch robotnik_gazebo_ignition spawn_robot.launch.py \
  robot:=rbwatcher x:=-19 y:=6 run_rviz:=true

Additional tweaks:

• Toggle RViz autostart via run_rviz:=false.
• Adjust initial pose with x/y parameters.
• Spawn extra robots by providing a unique robot_id so that namespaces stay isolated.

ros2 launch robotnik_gazebo_ignition spawn_robot.launch.py \
  robot:=rbwatcher x:=-19 y:=4 robot_id:=robot_2 run_rviz:=false

⚠️ Namespace reminder: All subsequent command snippets assume the default namespace /robot. If you launch with a custom robot_id such as robot_2, replace /robot with your namespace (for example /robot_2) in every topic or action name that follows.

---

Task 2 – Controllers and Sensors

Inspect command topics and sensor streams:

• Base odometry:

  ros2 topic echo /robot/robotnik_base_control/odom

• PTZ RGB-D camera (joint trajectory controlled):

  ros2 run rqt_image_view rqt_image_view /robot/top_ptz_rgbd_camera/color/image_raw

• Front RGB camera:

  ros2 run rqt_image_view rqt_image_view /robot/front_rgbd_camera/color/image_raw

• 3D LIDAR point cloud (visualize in RViz):

  

• IMU stream:

  ros2 topic echo /robot/imu/data

---

Task 3 – Basic Control

• Teleoperate the base using keyboard commands:

  ros2 run teleop_twist_keyboard teleop_twist_keyboard \
    --ros-args -r cmd_vel:=/robot/robotnik_base_control/cmd_vel -p stamped:=true

• Alternatively, drive from RViz using the teleop panel:

  

• Command the PTZ camera with the joint trajectory controller plugin:

  ros2 run rqt_joint_trajectory_controller rqt_joint_trajectory_controller --ros-args --remap __ns:=/robot

  

---

Task 4 – Mapping

Context: Nav2’s 2D planners and slam_toolbox consume sensor_msgs/LaserScan. The helper launch files flatten the robot’s 3D point cloud into a planar scan so SLAM and costmaps stay happy.

1. Project the 3D LIDAR into a planar scan:

   ros2 launch robotnik_simulation_bringup laser_filters.launch.py

   

2. Run slam_toolbox and drive the robot to build a 2D map:

   ros2 launch robotnik_simulation_localization mapping_2d.launch.py

   

3. Save the resulting occupancy grid:

   ros2 run nav2_map_server map_saver_cli -f ~/map

   

---

Task 5 – Localization & Navigation

Context: NavigateThroughPoses is a fire-and-forget mission—Nav2 visits each pose once and exits. FollowWaypoints suits patrols: it loops, tracks progress, and plays nicely with higher-level supervisors that pause/resume routes.

Localization

ros2 launch robotnik_simulation_localization localization.launch.py

Set the initial pose in RViz (2D Pose Estimate) and validate pose tracking while teleoperating the robot.

Navigation

Launch Nav2:

ros2 launch robotnik_simulation_navigation navigation.launch.py

Send goals via RViz or CLI:

ros2 action send_goal /robot/navigate_to_pose nav2_msgs/action/NavigateToPose '{
  "pose": {
    "header": {"frame_id": "robot_map"},
    "pose": {
      "position": {"x": 1.5, "y": 0.5, "z": 0.0},
      "orientation": {"x": 0.0, "y": 0.0, "z": 0.0, "w": 1.0}
    }
  }
}'

Waypoint following examples:

Navigate Through Poses

ros2 action send_goal /robot/navigate_through_poses nav2_msgs/action/NavigateThroughPoses '{
  "poses": [
    {
      "header": { "frame_id": "robot_map" },
      "pose": {
        "position": { "x": 0.0, "y": 0.0, "z": 0.0 },
        "orientation": { "x": 0.0, "y": 0.0, "z": 0.0, "w": 1.0 }
      }
    },
    {
      "header": { "frame_id": "robot_map" },
      "pose": {
        "position": { "x": 6.0, "y": 0.0, "z": 0.0 },
        "orientation": { "x": 0.0, "y": 0.0, "z": -0.707, "w": 0.707 }
      }
    },
    {
      "header": { "frame_id": "robot_map" },
      "pose": {
        "position": { "x": 6.0, "y": -6.0, "z": 0.0 },
        "orientation": { "x": 0.0, "y": 0.0, "z": 1.0, "w": 0.0 }
      }
    },
    {
      "header": { "frame_id": "robot_map" },
      "pose": {
        "position": { "x": 0.0, "y": -6.0, "z": 0.0 },
        "orientation": { "x": 0.0, "y": 0.0, "z": 0.707, "w": 0.707 }
      }
    }
  ]
}'

Follow Waypoints

ros2 action send_goal /robot/follow_waypoints nav2_msgs/action/FollowWaypoints '{
  "number_of_loops": 3,
  "poses": [
    {
      "header": { "frame_id": "robot_map" },
      "pose": {
        "position": { "x": 0.0, "y": 0.0, "z": 0.0 },
        "orientation": { "x": 0.0, "y": 0.0, "z": 0.0, "w": 1.0 }
      }
    },
    {
      "header": { "frame_id": "robot_map" },
      "pose": {
        "position": { "x": 6.0, "y": 0.0, "z": 0.0 },
        "orientation": { "x": 0.0, "y": 0.0, "z": -0.707, "w": 0.707 }
      }
    },
    {
      "header": { "frame_id": "robot_map" },
      "pose": {
        "position": { "x": 6.0, "y": -6.0, "z": 0.0 },
        "orientation": { "x": 0.0, "y": 0.0, "z": 1.0, "w": 0.0 }
      }
    },
    {
      "header": { "frame_id": "robot_map" },
      "pose": {
        "position": { "x": 0.0, "y": -6.0, "z": 0.0 },
        "orientation": { "x": 0.0, "y": 0.0, "z": 0.707, "w": 0.707 }
      }
    }
  ]
}'

---

Task 6 – Perception

Start the simple person detector package:

ros2 launch simple_person_detector simple_person_detector.launch.py

Monitor detection results:

ros2 topic echo /person_detector/detected
ros2 topic echo /person_detector/detection_array

---

Task 7 – PTZ Person Tracking

Bring up the PTZ tracker node:

ros2 launch ptz_tracker ptz_tracker.launch.py

Send a goal to start tracking:

ros2 action send_goal /ptz_tracker/start_tracking ptz_tracker_interfaces/action/TrackTarget '{start: true}'

The PTZ head attempts to keep the detected person centered until the action is canceled or detections stop.

---

Task 8 – Autonomous Inspection with Behavior Trees

Context: You could script the mission in one giant node, but Behavior Trees keep patrol, detection, tracking, and recovery logic modular. They make it easier to visualize state transitions, add fallbacks, and recover from failures.

The example tree implements a patrol-and-track policy:

1. Patrol a navigation route using Nav2 waypoints.
2. Interrupt patrol when a person is detected.
3. Delegate to the PTZ tracker while detections persist.
4. Resume patrol after the person leaves the scene.

Start the action server:

ros2 launch rbwatcher_behaviors behavior_tree_action_server.launch.py

Submit the default mission:

ros2 action send_goal /execute_behavior_tree rbwatcher_behaviors/action/ExecuteBehaviorTree '{
  target_waypoints: [
    { header: { frame_id: "robot_map" }, pose: { position: { x: 0.0, y: 0.0, z: 0.0 }, orientation: { x: 0.0, y: 0.0, z: 0.0, w: 1.0 } } },
    { header: { frame_id: "robot_map" }, pose: { position: { x: 6.0, y: 0.0, z: 0.0 }, orientation: { x: 0.0, y: 0.0, z: -0.707, w: 0.707 } } },
    { header: { frame_id: "robot_map" }, pose: { position: { x: 6.0, y: -6.0, z: 0.0 }, orientation: { x: 0.0, y: 0.0, z: -0.707, w: 0.707 } } },
    { header: { frame_id: "robot_map" }, pose: { position: { x: 0.0, y: -6.0, z: 0.0 }, orientation: { x: 0.0, y: 0.0, z: -0.707, w: 0.707 } } },        
  ],
  waypoint_loops: 1,
  waypoint_start_index: 0,
  tree_xml: "",
  tree_path: "config/default_tree.xml"
}'

The tree_path can point to custom XML if you author alternative mission plans in Groot.

---

Open Exercises

Ready to extend the workshop on your own? These drills progress from quick Behavior Tree tweaks to full autonomy experiments.

Level 1 – Behavior Tree Tweaks

• Inspection Patrol Remix: Edit config/default_tree.xml in Groot2 so FollowWaypoints alternates between two inspection points instead of looping four. Bonus: wrap each waypoint with a Delay decorator (10 s) to simulate on-site inspection time.
• Detection Hold-Off: Keep PTZ tracking while detections persist, but only resume patrol after IsPersonDetected has been false for 15 s. Combine timing decorators (for example Delay or Timeout) with a RetryUntilSuccessful guard to enforce the cool-down period.

Level 2 – New Behavior Tree Nodes

• Fixed PTZ Inspection: Implement a MovePTZToPreset SyncActionNode that drives the PTZ trajectory controller to a preset (pan = 1.0, tilt = –0.5) and holds for 5 s. Insert it into the main tree so the robot pauses at Waypoint 1, executes the preset scan, then advances to Waypoint 2.

Level 3 – Perception-Driven Logic

• Range-Aware Detection: Use /person_detector/detection_array to halt the patrol only when a detection is within 5 m. Extend the Bool condition or craft a new BT condition that checks the reported range before interrupting.

Level 4 – Navigation & Simulation Challenges

• Dynamic Obstacles: Spawn moving obstacles in Gazebo and rerun the patrol. Tune robotnik_simulation_navigation/config/nav2_params.yaml (for example controller_server.max_vel_x or global_costmap.inflation_radius) to make the robot more cautious or assertive and observe the change.

---

Happy patrolling! For issues or PRs, open a ticket on the repository and include environment details plus relevant logs.