costmap | Repository for costMAP software

I'm dealing with ros-navigation and its integrated layered costmaps. I have a static map in which obstacles from an occupancy grid need to be inserted. Since Obstacle Layer only can handle specific data (pointclouds from laser scanners etc.) and is apparently not able to handle a occupancy grid as input, I decided to write a custom layer which takes an occupancy grid and using the marking and clearing function from the occupancy grid to add obstacles and/or free space to the master grid. While running, there are no exceptions thrown. I figured out that the subscriber successfully receives the occupancy grid and enters the callback which at least runs through to the end. However, the updateCost-function (which apparently is responsible to add the modifications to the master grid) is never called (the ROS_INFO never throws its message). Therefore, no local map is generated, which causes RVIZ to throw a "No map received"-warning.

Any ideas of what is wrong with my layer?

The source-code of the custom layer:

I'm stuck with the following problem:

I have an occupancy grid based on information about the environment received from a camera plugged on my robot. Since the robot is supposed to move around in unknown territory, there is no "global costmap" in its original context. We use the information received from the camera as the "global costmap". Currently, the global costmap is configured as a static_layer that is linked to the base_link-frame (I need it to be a static_layer as it has multiple cost-values and only the static_layer with its "trinary_costmap"-parameter has the option to map intermediate cost_values range 0-255). Due to it is linked to the base_link-frame, it permanently moves along with the frame. But since the costmap is updated with a frequency of 2 occupancy grids per second, this permanent movement falsifies the costmap-information (since it's supposed to be a static map until the new occupancy grid gets published).

My idea was to create a new frame according to the ros tutorial for adding a new dynamic frame, which takes the current position of base_link and updates its position with a specific rate.
Therefore, my source-code looks like this:

I'm working on a workaround for the problem of this thread.

Since none of the existing costmap2d-layers appears to allow the usage of the full range of values (0-255) I used the ros-tutorial to create a custom layer. Therefore, I just used the source-code of the static_layer plugin and modified the interpretValue - function in order to map the value (which is due to the used occupancy grid between -1 and 100) to the full range of the layer (which should be 0-255). I integrated my custom plugin into the global_costmap_params.yaml and the system appears to properly load the plugin (at least there are no further errors or warnings that it couldn't be loaded).

The problem is: In RVIZ the global costmap - section throws a warning which says "No map received" (The Topic is '/move_base/global_costmap/costmap' which works fine when static_layer is set as plugin). As a result of that, I can only see the coordinate system, but no map.

I'm using ROS Melodic.

Plugin source code (occgrid_to_costmap_layer.cpp):

I am using RP-Lidar /scan topic together with move_base from navigation stack. Although the obstacle layer is parameterized to get the LaserScan type data from /scan topic, I am recieving the message that "Recover behavior will clear layer 'obstacles'". I would like to mention that the UniTest over /scan topic and /odom topic are working fine. Thus in my RVIZ, the obstacle are not shown neither the planner takes them into account to prevent a collision. For clarity here is my common config file:

I have a relatively simple issue with the initialization of the robot pose in navigation stack. After doing a SLAM in the environment and saving the static map from map_server with desired resolution and size, and after finally making it available from static map server, I can not properly initialize my robot pose. Although, the initialization is possible with RVIZ (2D pose estimation option), it is inaccurate obviously since I am doing it by hand. This has a second draw back namely, when I am using a position (/move_base_trajectory topic) tracking instead of velocity tracking (/cmd_vel topic), there is an offset of the current position which I should manually compensate in my low-level controller. Is there any other proper (automatic) pose initialization approach?
PS: I am using Lidar /scan topic to feed costmap functionalities. I have also tested the initilization of the robot with hector_slam stack which pretty accurate however, with that I am getting a second map which interferes with the first one and pose association does not work. There should be a more proper approach I guess. I am loading my static map within the launch file:

I am working with move_base from navigation stack. However, I am getting the warning that

"local_costmap: preHydro parameter "static_map" unused since "plugins" is provided"

In terms of costmap definition here are the common and local config files I have been using: