ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | ros @ Robotics Stack Exchange
Ask Your Question
0

what will be the best sensor suite for an indoor robot ?

asked 2015-06-12 11:50:50 -0500

dinesh_sl gravatar image

Dear all ,

I am asking this to get some help from experienced roboticists here. Forgive me if this is not the correct place to ask this question. We are planning to implement autonomous indoor robot for a large indoor environment for a university project. Is LiDAR a must for this kind of a robot ? or could we be satisfied with an RGB-D sensor (addition to IMU + odometry)?

Your ideas are really appreciated. Thanks.

edit retag flag offensive close merge delete

3 Answers

Sort by » oldest newest most voted
1

answered 2015-06-12 23:52:02 -0500

updated 2015-06-13 23:08:46 -0500

First you should have a clear definition on your large indoor environment because range sensors like LRF or RGB-D sensor have limitation on distance measurement. For example, Kinect has practical ranging limit around 0.8 – 4 m.

Camera does not have the limitation, but it lacks the absolute info about scale. Also, related tools are less. LSD-SLAM(Large Scale Direct SLAM) is a good one you might want to take a look. But as far as I know, navigation with monocular is still a problem.

http://vision.in.tum.de/research/lsdslam

Hope this helps.

edit flag offensive delete link more

Comments

Thank you very much. Environment will be populated with still objects like in a library. So range problem will not be a much problem. Because of the cost we are planning to use RGB-D sensor information to laser scan because we are given a 2D map of the environment. Is this a suitable method ? thanks

dinesh_sl gravatar image dinesh_sl  ( 2015-06-13 23:51:22 -0500 )edit

Probably works. With pointcloud_to_laserscan ( http://wiki.ros.org/pointcloud_to_las... ), you can use the navigation package to build your exploring robot system. If your task is patrolling, SMACH can provide a smart solution to handle the whole task.

Po-Jen Lai gravatar image Po-Jen Lai  ( 2015-06-14 00:57:30 -0500 )edit
1

Or you avoid the costly pointcloud by using depthimage_to_laserscan

Humpelstilzchen gravatar image Humpelstilzchen  ( 2015-06-22 02:51:16 -0500 )edit

Overall, RGB-D sensor is a good idea since you can use the RGB image or you can transform depth to laser according to your scenario. Though the accuracy might not be as good as LIDAR.

Po-Jen Lai gravatar image Po-Jen Lai  ( 2015-06-22 03:15:14 -0500 )edit
0

answered 2015-06-21 10:53:00 -0500

duck-development gravatar image
  • If you have daylight an RGB-D sensor is not the best choise.
  • The Computation offer with a RGBD Sensor ist mutch bigger then with the Serial Data Stram of an LIDAR Sensor.
edit flag offensive delete link more
0

answered 2015-06-12 21:51:01 -0500

wsAndy gravatar image

I wonder if this video is similar to what you say .https://www.youtube.com/watch?v=NmnepqAQIlk rgbdslam is also a good choose.

edit flag offensive delete link more

Comments

thanks for your idea. It is very much similar. We are given a 2D map of the environment. Will converting rgb-d data to a laser scan works rather than making a map using kinect ?

dinesh_sl gravatar image dinesh_sl  ( 2015-06-13 23:43:22 -0500 )edit

If you use kinect, you can create a map in 3D, but it requires a higher performance for your computer. Only use kinect can create a map directly too.

wsAndy gravatar image wsAndy  ( 2015-06-15 03:05:59 -0500 )edit

Question Tools

2 followers

Stats

Asked: 2015-06-12 11:50:50 -0500

Seen: 1,179 times

Last updated: Jun 21 '15