NVIDIA-ISAAC-ROS / isaac_ros_map_localization

NVIDIA-accelerated global localization
https://developer.nvidia.com/isaac-ros-gems
Apache License 2.0
57 stars 6 forks source link

Lack of Precision in Isaac ROS Ocupancy Grid Localizer #26

Closed pabloinigoblasco closed 10 months ago

pabloinigoblasco commented 11 months ago

Issue Description: Insufficient precision in isaac_ros_occupancy_grid_localizer for the Carter robot, leading to inconsistent pose estimation between consecutive time steps. Consequence: Results in a visually jarring "jump effect" in position estimation during radial motion execution in RViz. Impressions: isaac_ros_occupancy_grid_localizer might be misconfigured for the current robot/lidar setup. The new FOV in the 2023 version is suspected to be a contributing factor, as the localization worked in 2022 but is problematic in 2023. Evidence:

VIDEO: https://www.dropbox.com/scl/fi/hqhltqwemr0nujmjuznc5/2023-11-06-10-30-55_sm_isaac_exp_grid_loc_problem.mp4?rlkey=ilv0qvvz1n59jpab95rn6ck7s&dl=0

image

gordongrigor commented 11 months ago

isaac_ros_occupancy_grid_localizer is designed to provide an initial global localization. It's not a continuous localization package.

tugbakara commented 7 months ago

In the documentation page of this package there is a note about the continuous localization. Am I wrong or what do you mean for this note? I am confused.

HappySamuel commented 3 months ago

In the documentation page of this package there is a note about the continuous localization. Am I wrong or what do you mean for this note? I am confused.

Agree with @tugbakara , I thought that /flatscan_localization is used for continuous localization as well. As it's written that whenever flatscan is sent to here, localization will be triggered.