To train/test different algorithms for their obstacle avoidance capabilities, it would be helpful to have obstacles in Gazebo.
It would be good to have a node that places random objects (cubes, cylinders,...) along the racetrack. It would also be useful if the node could reposition them upon receiving a message. That way the obstacles can be rearranged by learning algorithms like Q-learning or evolutionary algorithms.
There is a working version on the branch marian-obstacles, but is has some performance problems in Gazebo. It seems to be a problem with Gazebo, see here.
To train/test different algorithms for their obstacle avoidance capabilities, it would be helpful to have obstacles in Gazebo.
It would be good to have a node that places random objects (cubes, cylinders,...) along the racetrack. It would also be useful if the node could reposition them upon receiving a message. That way the obstacles can be rearranged by learning algorithms like Q-learning or evolutionary algorithms.