[ros-users] Waypoint navigation
s34.martin at connect.qut.edu.au
Fri Nov 12 04:07:05 UTC 2010
To be honest I am not sure you would gain very much from the Nav stack in your case. I think it should be possible, you would need to broadcast a transform from you current localisation system and possibly publish blank costmaps. I think the blank costmaps would be the easiest work around given the very integrated nature of the Nav stack.
So yes it could be done but you are then constrained to work with the current framework if you want to write your own planning or control algorithms. This can be a bit restrictive if you are just trying to get stuff up and running but it does mean you can use all of the existing work and extend the robots capabilities such as using other sensors if they are incorporated.
There isn't a lot of algorithms implemented for the Nav Stack at the moment. (In the stable release anyway, I think there might be some planners being implemented in the experimental stuff.) I am working on a few other path following and planning algorithms when I have time but the current state is a reliable solution for the PR2 not an all purpose solution for path planning on every vehicle. So it really depends where you want to go with it in the future as more functionality will become available but for the moment it would probably be easier to do your own implementations.
Hope this was helpful.
From: ros-users-bounces at code.ros.org [ros-users-bounces at code.ros.org] on behalf of Mike Purvis [mpurvis at clearpathrobotics.com]
Sent: Friday, 12 November 2010 7:04 AM
To: User discussions
Subject: [ros-users] Waypoint navigation
We do a lot of in-house testing of our platforms, for reliability, battery life, etc. The basic pattern for these tests is to set the vehicle with a bunch of waypoints and have it navigate a circuit for a few hours. Currently, waypoint management is through a custom web gui, and all the planning and so-on is very simple, just one goal at a time, point and shoot, etc.
We're looking at increasing the complexity on the control side (for example, using a carrot-based planning approach), and I'm just wondering if there are published modules out there which we could be integrating with, rather than continuing to build on our own stuff. Part of the idea, of course, would be to open some of it up eventually, but that does no good if it doesn't integrate with any of the conventions that are out there.
At the time that we were setting this all up in June, we didn't have a LIDAR or a map server or any of the baseline requirements for Eitan's Navigation stack. And given that we wouldn't be using any of the SLAM/exploration-type functionality, it still seems like the wrong tool for the job in our case. Is it, or is there some of the planning and path management stuff that would indeed meet our needs, and be usable outside of the assumptions regarding mapping, obstacle avoidance, sensors, etc? If not the Nav stack, are there other components that address these needs?
More information about the ros-users