costmap mapping the map without noise

Question

hello everyone,I am new in ROS and I'am studing with explore package. wiki.ros.org/explore. I run the code without problems. But When I use the rviz to see the explored map,I find that after the costmap got laser scan's data, as soon as the obstacle is detected, it has a probability to be a Free Space while if the truth is Obstacle, that made the wall(in the rviz ) it detected has some sawtooth shape. So I want to ask that how can the map mapping the stage's map exactly and I want it mapping without noise,which parameters should I change?

Thank you every much !

---

Originally posted by physicsex on ROS Answers with karma: 11 on 2014-03-19

Post score: 0

---

Original comments

Comment by demmeln on 2014-03-23:
You will have to be a bit more specific than that. What exactly are you doing? Which package are you using? What commands are you executing? What is happening? What would you like to happen?

Answer 1

Not sure if I understand you correctly, but the sawtooth shape/pixelization of the costmap happens because space is represented by grid cells with a fixed cell size (typically 5cm). What you could try is reducing the cell size, but that will have various side-effects. The point of the cost map isn´t so much good looks but rather well working obstacle avoidance of the robot. You could try writing your own improved visualization if you absolutely require nicer visualization.

---

Originally posted by Stefan Kohlbrecher with karma: 24361 on 2014-03-24

This answer was ACCEPTED on the original site

Post score: 2

---

Original comments

Comment by physicsex on 2014-03-25:
Thank you very much for the explaining, but what I mean is not for the reason of the pixelization problem, I upload a picture for this problem https://docs.google.com/document/d/1h9QVhTbo5x0NvXeEBGWSJZVCrNJ1-ZXSdpWzx7Hx6vM/edit?usp=sharing you can see on the walls there are some white blocks which I want to erase .

Comment by Stefan Kohlbrecher on 2014-03-27:
Those are artifacts due to noise in both sensor data and state (i.e. pose) estimation as well as discretization of the environment. There are basically two things you could look at:

• Postprocess map data using various computer vision approaches
• Change the actual map generation process to try to get rid of these artifacts. Both will be pretty involved if you want to achieve really good results, so I´d recommend leaving things as they are :)

Comment by physicsex on 2014-04-05:
so the map generation process comes from the code of the costmap? does costmap used gmapping or other estimation methold of the robot?Then the sensor data come from the Stage and it don't have any noise. Thank you very much!