Robotics StackExchange | Archived questions

Cspace in navigation stack?

Hi,

Where does the ROS navigation stack consider the configuration space of the robot for circular mobile robots, such as the Turtlebot 2?

I am looking through the wiki, but I cannot find much about it. The global planner seems to plan on a workspace occupancy grid. If that is the case, then does the global planner do any kind of obstacle avoidance (even discarding nodes that are occupied locations)? Or does the global planner plan without considering obstacles and the local planner handles all obstacle avoidance?

Asked by sterlingm on 2018-09-07 16:22:41 UTC

Comments

Answers

The map is the configuration space of traversable space in a 2D fixed-map mobile robotics application. It will visit the nodes belonging to occupied cells and look at the cost and using standard graph search techniques throw out from the path as untraversable space.

The local planner does take care of some obstacle avoidance but assuming a perfect map to perfectly represent your environment, it shouldn't have to do anything in theory, but not in practice or any amount of non-gazebo reality. The local planner also is what produces velocity commands to the wheels rather than just a list of ordered cells to visit.

Asked by stevemacenski on 2018-09-07 18:20:56 UTC

Comments

Do you mean that the occupancy grid published on '/map' represents the configuration space of the robot (for 2D applications)? So the obstacle pixels found are already inflated based on the robot's size and shape?

Asked by sterlingm on 2018-09-09 12:34:56 UTC

The configuration space is represented with the "inflated" portions of your costmap. All the lethal obstacles have an additional lethal cost added to them based on the inscribed radius of your robot.

Asked by David Lu on 2018-09-19 10:50:47 UTC

Comments