ros-follow-object

ROS follow object example

This is an example of ROS driver being run as an OAK 4 app together with a custom follow_object ROS2 node. The app launches depthai_filters spatial bounding boxes (/spatial_bb) and then uses spatial detections to follow a selected object. Object selection is done by filtering marker text in code with self.OBJECT_MARKER_TEXT (default: "person"). The node computes linear/angular velocity commands from the selected marker position and publishes them to /cmd_vel. It is based on ROS2 Kilted, but you should be able to subscribe to topics in other distributions such as Humble or Jazzy.

Demo

Prerequisites

Before you begin, ensure you have the following installed on your host machine:

• ROS2 (Humble, Jazzy, or Kilted)
• Rviz2

Setup Instructions

1. Install the required ROS packages:

   sudo apt update
   sudo apt install ros-$ROS_DISTRO-rviz2

2. Source your ROS2 environment:

   source /opt/ros/$ROS_DISTRO/setup.bash

Running the Example

1. Launch the OAK 4 app:

   cd <path_to_this_example>
   oakctl app run .

2. In a new terminal, source your ROS2 environment and inspect topics:

   source /opt/ros/$ROS_DISTRO/setup.bash
   ros2 topic list

3. Optionally monitor output commands:

   ros2 topic echo /cmd_vel

Changing the Target Object

To follow another detected object class, edit:

• apps/ros/ros-follow-object/src/follow_object/follow_object/follow_object_node.py

and change:

self.OBJECT_MARKER_TEXT = "person"

to the desired marker text label (for example "bottle" or "car"), then rebuild/restart the app.

Visualizing Data in Rviz

1. In Rviz, add a "Marker" display.
2. Set the topic to /spatial_bb.
3. Optionally add other displays from depthai ROS topics if needed.

Package deployed on an iRobot Create 3

This package can be adapted to any robot that listens to the /cmd_vel topic for motion control. We deployed it on an iRobot Create 3; read more in this blog post.

Troubleshooting

If you encounter issues with topic names or data types, verify that your ROS2 distribution matches the one used in the example. You may need to adjust topic names or data types accordingly.

sudo apt install ros-$ROS_DISTRO-rmw-cyclonedds-cpp
export RMW_IMPLEMENTATION=rmw_cyclonedds_cpp