ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | ros @ Robotics Stack Exchange
Ask Your Question

marcusrm's profile - activity

2013-05-29 09:39:48 -0500 commented answer Navigating with a differential drive rectangular robot

We think that the "forward_point" parameter is probably the best improvement for us because it keeps our robot's heading more consistent.

2013-05-29 09:38:13 -0500 commented answer Navigating with a differential drive rectangular robot

Our problem has since been resolved (mostly...). We think that our robot was too long and rectangular to use the trajectory_planner, however we were able to fix some bugs that were preventing us from using the DWA planner which works much better.

2013-05-29 09:34:51 -0500 received badge  Famous Question (source)
2013-05-29 09:33:35 -0500 received badge  Good Question (source)
2013-05-27 15:39:33 -0500 received badge  Favorite Question (source)
2013-05-25 09:27:15 -0500 received badge  Notable Question (source)
2013-05-25 04:16:15 -0500 received badge  Nice Question (source)
2013-05-25 03:18:33 -0500 received badge  Popular Question (source)
2013-05-24 23:11:59 -0500 received badge  Student (source)
2013-05-24 10:56:23 -0500 asked a question Navigating with a differential drive rectangular robot

Currently we are using the Navfn global planner with the Trajectory_Rollout local planner on our differential-drive robot that is approximately 1 meter by 0.6 meters (with the center of the differential-drive being approximately 0.35 meters from the back of the robot). You can see a picture of our robot here:

image description

So far we have had limited success with our path planning. At the moment we believe that our biggest problem is that the robot will attempt to keep its center point (the center of the wheelbase) on the global path at all times, and if it moves slightly off the path it will end up spinning (usually a multiple of 360 degrees) until it merges back onto the path where it left off with the correct heading. (We have already tried all suggestions from this thread those linked on the same page)

We have been reading old forum messages and we think that our problem may be that the Navfn planner does not give pose requirements to local goals, meaning that when we reach curves on our path the robot will come in with the wrong heading and start its spinning problem. We wanted to use the pose_follower based on dialogue from this archive but it gave no messages and didn’t move when we ran it. We also tried the sbpl_lattice_planner to no avail.

We are wondering whether or not a global/local planner has been developed that works well with rectangular shaped robots. From our experience and what we’ve read from other users with similarly shaped differential drive robots, tuning the parameters on our current planners is not enough to get a reliable solution. We looked at the development on the PR2 for cart pushing, but we were unsure if we should attempt to make our robot "look" like a smaller square robot pushing a similarly shaped square cart.

Your feedback is greatly appreciated.


Here are our parameters that we're currently using:

parameters directory

2013-05-23 21:52:08 -0500 received badge  Great Answer (source)
2013-05-23 21:52:08 -0500 received badge  Guru (source)
2013-05-23 05:57:42 -0500 received badge  Good Answer (source)
2013-05-23 05:57:42 -0500 received badge  Enlightened (source)
2013-05-23 05:27:36 -0500 received badge  Supporter (source)
2013-05-23 05:26:59 -0500 received badge  Nice Answer (source)
2013-05-23 05:26:24 -0500 received badge  Teacher (source)
2013-05-23 05:25:42 -0500 answered a question gmapping missing spots

For the SICK LMS lidar you can set the "use_rep_117_" global parameter to "false" to force it to report readings in legacy format instead of using Inf/NaN. It seems that the gmapping package rejects Inf/NaN values.

Use this information in combination with the maxRange and maxUrange suggestion that Ben has offered.