Running a trained model on a real Turtlebot

[LevinTamir]

Is it possible to run a trained model from Gazebo on a real turtlebot using the dqn pkg availble here? If so, what are the nodes that need to be running?
If not, what modification need to be done?
Thanks

[LevinTamir]

I'm running ROS2 Humble & Gazebo Classic, I fogot to mention.

[GyuH13]

Hello @LevinTamir
During training, 24 LiDAR data points from the front 180 degrees, along with the distance and angle to the goal, are used.
To apply this directly to a real robot, you will need to modify the /scan topic callback to extract the 24 front-facing data points.
Additionally, you will need to manually handle the role of the dqn_gazebo node, which includes generating and sending the goal pose.
Significant source code modifications are likely required to use this in a real-world environment.

[LevinTamir]

Thank you for the fast response!
Any examples availble?

[chan-1207]

Additional feature development is not possible in the near future due to scheduling constraints. For application to a real robot, the most important changes are the nodes that replace the role performed by dqn_gazebo.py and the filtering of lidar data from the real TurtleBot3.