mattiasbäx's profile - activity

I might have figured it out, posting my findings here if anyone browses this thread:

1) I set the obstacle layer of the laser scanner to "enabled: false". This allows me to approach objects without planner stopping me. 2) Setting the min_vel_x to max_vel_x but switching sign solved this. 3) No, the parameters are updated when dynamically reconfigured.

Hello ROS-community,

I have a problem you might help me figure out. I am using a husky to pick up certain objects, let's say these objects are boxes. I have a breadcrumb following navigation node that communicates with move_base to maneuver the environment and approach the box. The problem is that if i want to approach the box the husky will treat the box as an obstacle, thus trying to avoid it. Is it possible to "shut off" the obstacle avoidance, pickup the box, and power it on? I was reading this thread:

http://answers.ros.org/question/12276...

And thinking there might be a parameter i could reconfigure during runtime. Another problem i have is that i want to be able to take the fastest path if in order to align the husky to the box properly. If i give a goal point to move_base that's behind the husky, i want it to go backwards. But at the same time if the goal point is in front of the husky, it should move forward. Is this a possible behavior using dynamic reconfigure?

Another think with dynamic reconfigure as the thread above: Is there a need to call some sort of update, in order for the new parameters to take place? I made a simple node like the thread above to reconfigure some move_base parameters but when i read them from the terminal they hadn't changed.

TLDR: 1) Is it possible to "shut off" and "power on" obstacle avoidance during runtime? 2) Is it possible to have a behaviour that takes fastest path possible, which goes backwards if needed, and switch to "normal" maneuvering behaviour during runtime? 3) Do I need to call some sort of update for my dynamically reconfigured parameters to take place?

Thanks in advance! Wbr Mattias

I have managed to create a node that publishes fake sensor data to an obstacle layer i created myself in the global of move_base. But I have no idea how to clear this layer to completely remove all the virtual walls.

if it is not possible to do so, the robot should request another path and/or area. Is it possible though, to clear the previous virtual walls once a new set of waypoints and a new "corridor" is recieved?

I don't think that the only path being through the virtual walls and the robot not taking it is a problem, it's actually the behavior I'm looking for. The robot is supposed to be a system in a system, where the decision of "move through these waypoints in this area" is made on a higher lvl, and ...

Thank you Daenny for a more elaborate description on solution 2. Would it be possible for you to give a step by step breakdown of problem one, as you did in your last comment of the post you linked?

Thank you both. Another question: I want to do this every time a new set of waypoints are given to the robot. Once a set of waypoint is published, the map is updated with the virtual corridor accordingly. Is this still possible? Because editing the static map is just before runtime, no?

Thank you Stefan, this is very useful! How will these different solutions affect the localization with for example amcl? Is it possible to add obstacles to the costmap that the robot avoid, but don't localize against (since they don't exist)?

Hello ROS community!

I just got into ROS through a project I'm participating in at my University. My job is to navigate the husky given a set of breadcrumbs (waypoints) and a zone (like a "virtual corridor") where it's OK to maneuver in.

Whilst playing around with the husky gazebo simulation and investigating how the move_base local and global planner work, I thought of an idea. Would it be possible to add a "virtual corridor" to the global occupancy grid cost map, which would make the husky think it was trapped and not navigate outside of the corridor, but still navigating through the breadcrumbs and avoiding obstacles inside? My hopes is that the laser distance sensor could still see through this virtual wall in order to navigate itself in the world.

Or do you know of any more convenient way to do this?

Thanks in advance, wbr

EDIT: Basically what I'm asking is: Is it possible to add area restrictions on the path planner in order to keep the robot inside a certain region while going to goal?