0
$\begingroup$

Rosanswers logo

I am building a robot and implementing autonomous navigation using gmapping, amcl, and move_base. From my understanding the amcl package compares the map to lidar data (sensor_msgs/LaserScan) to update the map->odom transform. I also need to implement a publisher to calculate the odom->base_link transform from encoder data. Both map->odom and odom->base_link are attempting to localize the robot. This seems redundant.

Is it possible to localize the robot with only LaserScan data by publishing a direct map->base_link transform? Is the odom->base_link transform only necessary when building a map (using gmapping)?


Originally posted by fruitbot on ROS Answers with karma: 19 on 2022-01-20

Post score: 0

$\endgroup$

1 Answer 1

0
$\begingroup$

Rosanswers logo

According to ROS REP-105, the base_link is a rigid frame attached to the robot whereas both odom and map frames can be non-rigid. The difference between them is that the odom frame may drift over time (which makes it useless for long term reference) whereas the map frame should not drift over time but may be non continuous (may cause discrete jumps) in the short term.

This is why systems often use both frames for localization.

Is it possible to localize the robot with only LaserScan data by publishing a direct map->base_link transform?

Yes, of course!
Is it recommended? No. Nothing is stopping you from using it other than convention.


Originally posted by Akhil Kurup with karma: 459 on 2022-01-21

This answer was ACCEPTED on the original site

Post score: 1


Original comments

Comment by fruitbot on 2022-02-08:
"Is it recommended? No."

What is the downside to using only LaserScan for localization? I see packages exist that calculate odometry based only on lidar data. My guess is it is more computationally intensive to rely only on laser scan localization (as opposed to dead reckoning with encoders/accelerometer)?

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.