Driftbot — Autonomous Tidying Robot

A ROS 2 Jazzy + Gazebo Harmonic project. A four-wheeled differential-drive robot with a 3-joint magnetic arm picks up objects scattered around a simulated home and drops them in a collection bin.

Packages

driftbot_description

Robot definition. Contains:

• urdf/ — xacro files: base.xacro (chassis, 4 wheels), torso.xacro, arm_active.xacro (shoulder/elbow/wrist joints), arm_cosmetic.xacro, sensors.xacro (camera, GPU lidar, IMU), materials.xacro, inertia_macros.xacro, ros2_control.xacro
• urdf/robot.urdf.xacro — top-level xacro that includes everything plus Gazebo plugins: diff-drive, joint state publisher, joint position controllers (arm), detachable joint system (magnet pick/drop), triggered publishers (magnet_off fan-out)
• meshes/ — STL meshes for base, wheels, torso, arm links
• launch/view_robot.launch.py — RViz-only robot viewer
• config/ — RViz configs for different stages
• rviz/ — saved RViz layout

driftbot_gazebo

Simulation world. Contains:

• worlds/home.sdf — two-room home with walls, furniture, three pickable objects (toy_block_1, can, ball), and a collection bin
• models/ — SDF + DAE meshes for furniture (chairs, dining table, sofa, coffee table, side table, bookshelf, file cabinet, fridge) and the drop-off box
• launch/home.launch.py — starts Gazebo (ogre2), robot_state_publisher, spawns robot via ros_gz_sim create, fires initial magnet_off after 5 s

driftbot_task

Mission logic. Contains:

• driftbot_task/task_node.py — TaskNode class: odom/IMU/scan/camera subscriptions, world-frame pose (odom position + IMU heading), motion primitives (face_toward, drive_to, drive_straight, turn_by_imu, settle_heading, bug2_drive_to), arm staging, pick()/drop(), mission dispatch
• driftbot_task/gz_support.py — GzBridgeSupport node: relays joint_states_gz→joint_states, broadcasts odom→base_footprint TF, relays scan_raw→scan with lidar_link frame
• driftbot_task/missions/ — __init__.py (dispatcher), mission_01.py (straight-line block pick+drop), mission_02.py (Bug2 can+ball pick+drop)
• config/gz_bridge.yaml — 15 ROS↔Gazebo topic mappings
• launch/mission.launch.py — staggered launch of Gazebo, bridge, nodes, RViz
• rviz/mission.rviz — RViz layout with grid, robot model, TF, laser scan, camera

---

Dependencies

• ROS 2 Jazzy
• Gazebo Harmonic
• ros_gz_bridge, ros_gz_sim
• robot_state_publisher, tf2_ros
• rviz2, image_view
• Python 3

No Nav2, no OpenCV, no external planners.

---

Setup & Installation

1. Clone the repository

git clone <repo-url> ~/driftbot_assignment/driftbot_ws
cd ~/driftbot_assignment/driftbot_ws

2. Build the workspace

source /opt/ros/jazzy/setup.bash
colcon build --symlink-install

3. Update ~/.bashrc

Open ~/.bashrc in a text editor and add (or update) the following lines. If you previously had these pointing to a different folder (e.g. drift_assignment instead of driftbot_assignment), replace the old lines entirely — do not keep both.

source ~/driftbot_assignment/driftbot_ws/install/setup.bash
export GZ_SIM_RESOURCE_PATH=$GZ_SIM_RESOURCE_PATH:/home/$USER/driftbot_assignment/driftbot_ws/install/driftbot_description/share
export GZ_SIM_RESOURCE_PATH=$GZ_SIM_RESOURCE_PATH:/home/$USER/driftbot_assignment/driftbot_ws/src/driftbot_gazebo/models

Note: GZ_SIM_RESOURCE_PATH tells Gazebo where to find meshes and models. If this is stale or points to the wrong directory, Gazebo will fail to load all STL meshes and the robot will appear invisible in the simulation.

4. Apply the changes

source ~/.bashrc

5. Verify the resource path

echo $GZ_SIM_RESOURCE_PATH

The output should include paths under ~/driftbot_assignment/. If you still see the old folder name, check ~/.bashrc for duplicate or leftover entries and remove them.

---

Launch Files

driftbot_description — view_robot

Opens RViz with the robot URDF model for inspection. No simulation.

ros2 launch driftbot_description view_robot.launch.py

driftbot_gazebo — home

Starts Gazebo with the two-room home world, runs robot_state_publisher, and spawns the robot. Does not start any bridge or task nodes.

ros2 launch driftbot_gazebo home.launch.py

Optional args: spawn_x, spawn_y, spawn_yaw, spawn_z, software_gl.

driftbot_task — mission

The main launch. Starts everything in sequence:

1. 0 s — Gazebo world + robot spawn (includes home.launch.py)
2. 8 s — ROS↔Gazebo bridge (clock, odom, cmd_vel, joints, scan, arm, IMU, camera, magnet topics)
3. 12 s — gz_support node (joint relay, odom→TF, scan frame fix) + task_node (runs mission)
4. 15 s — RViz + camera viewer

ros2 launch driftbot_task mission.launch.py

Optional args: spawn_x, spawn_y, spawn_yaw, mission_id.

---

Missions

A single launch runs all phases in sequence:

Mission 1 — Pick block (straight-line) Drives south to toy_block_1, stops at 60 cm, picks up with magnetic arm, turns 180°, locks heading north, drives to drop-off point, turns -90° to face east, locks heading, drops into collection bin.

Mission 2 — Pick toy_block_2 + toy_block_3 (Bug2) Uses Bug2 obstacle-avoidance to navigate to toy_block_2, picks it up, Bug2 back to bin, drops. Then repeats for toy_block_3.

Mission 3 — Computer vision object detection pipeline TBD.

---

World Layout

Robot spawns at (7.5, 9.0) facing south (yaw = -π/2).

Object	Position
toy_block_1	(7.5, 5.0)
toy_block_2	(-5.0, 5.0)
toy_block_3	(-2.0, 2.0)
can	(1.5, 9.0)
ball	(1.5, 2.0)
cone	(-8.0, 3.0)
collection bin	(8.50, 9.0)
drop-off point	(7.5, 9.0)

---

References

• Rover STL meshes and URDF adapted from GGomezMorales/waver
• Arm URDF and meshes adapted from AL5D description
• Furniture models sourced/adapted from Gazebo Fuel, osrf/gazebo_models, and ipa320/cob_simulation