Carlos Miguel Correia da Costa

Good night, From the tests I made in the past, ICP typically works better: https://www.researchgate.net/publication/322664340_Map-Matching_Algorithms_for_Robot_Self-Localization_A_Comparison_Between_Perfect_Match_Iterative_Closest_Point_and_Normal_Distributions_Transform Have a nice day,

Good night :) The ICP algorithm does not require two point clouds. It only requires the [corresponding points between two point clouds](https://github.com/carlosmccosta/pcl/blob/master-all-pr/registration/include/pcl/registration/impl/icp.hpp#L192) to be [computed](https://github.com/carlosmccosta/pcl/blob/master-all-pr/registration/include/pcl/registration/impl/correspondence_estimation_lookup_table.hpp#L590). As such, the reference point...

Good afternoon, The [ICP](https://en.wikipedia.org/wiki/Iterative_closest_point) registration algorithms try to find the least squares minimization between the distances of the corresponding points [sensor data -> map]. This means that objects that are...

Good afternoon, I tested the guardian_config configurations in a [ML300 Rend Lake NUC](https://www.onlogic.com/ml300), equiped with a [2.80 GHz Intel Core i5-3427U](https://cdn.onlogic.com/pub/media/resources/spec-sheets/LogicSupply-ML300-SpecSheet.pdf), which had better performance when [compared with the Intel...

Hello :) The `update_navigation_map` in `guardian_config` was setup to update only the map for navigation. The goal was to use a static and clean map (without garbage / dynamic objects)...

In slam mode drl starts with no map and builds it over time when new sensor data arrives. But you can give drl an occupancy grid in the beginning for...

`reference_pointcloud_update_mode` specifies the modes of slam: - NoIntegration -> static map - FullIntegration -> all aligned data is used for updating the map - InliersIntegration -> only points close to...

Yes, when `publish_pose_tf_rate` is map, and since we had +- good odometry, I could configure pose_to_tf_publisher to republish the last odom -> map correction with an updated timestamp so that...

If you are planning on using slam, then you should probably use kd-trees instead of lookup tables, since a kd tree is usually faster to initialize.

Yes, you should just adjust the part relative to the correspondence_estimation_approach: https://github.com/carlosmccosta/dynamic_robot_localization/blob/melodic-devel/yaml/schema/drl_configs.yaml#L506 Just change it to: `CorrespondenceEstimation`.