This branch contains a ROS-implementation of Next-Best-Sense adapted for working on topological maps. It relies on the ROS package bayesian_topological_localisation for updating the belief over the objects tracked.
A video of the working principle of topoNBS combined with a topological particle filter for localizing human pickers in a polytunnel is hosted on YouTube.
This repository depends on the RASberry package which is not released yet. You are required therefore set up your own simulation before being able to use topoNBS.
This is a short description on how to install and run the components used for the topological localization presented above.
- Install the LCAS ROS software packages: https://github.com/LCAS/rosdistro/wiki
- Setup
topological_navigation: https://github.com/LCAS/topological_navigation/wiki/Quick-Topological-Navigation-in-simulation-tutorial - How to use
topological_navigation: https://github.com/LCAS/topological_navigation/blob/master/topological_navigation/README.md
The package is already installed if you have followed the previous step. Just follow the instructions to use: https://github.com/francescodelduchetto/topological_navigation/blob/master/bayesian_topological_localisation/README.md
In order to launch the Next-Best-Sense framework for navigating the robot, you can simply use the following command
roslaunch next_best_sense pure.launchIf you have a look at the launch file, there are some interesting parameteres which can be tuned. In particular,
<arg name="w_info_gain" default="0.00"/>
<arg name="w_travel_distance" default="0.2"/>
<arg name="w_sensing_time" default="0.1"/>
<arg name="w_battery_status" default="0.2"/>
<arg name="w_rfid_gain" default="0.5"/>are the five criteria currently implemented for choosing the next robot pose. A value in [0,1] can be assigned to each of them in order to express a relative preference among them. Details of how the criteria are combined is offered in the relative paper.
The following piece of code launch the topoNBS experiments for one or two RFID antennas (defined by the argument two_readers). It requires the package RFID in order to work with likelihood readings coming from the antennas.
<node pkg="next_best_sense" type="pure_navigation" name="nbs" output="screen" args="$(arg pure_args)" unless="$(arg two_readers)">
<param name="move_base_costmap_topic_name" value="/map"/>
<param name="move_base_costmap_updates_topic_name" value="/move_base/global_costmap/costmap_updates"/>
<param name="belief_map_srv_name" value="/thorvald/rfid_grid_map_node/get_rfid_belief"/>
<param name="fake_belief_map_srv_name" value="/thorvald/rfid_grid_map_node/get_rfid_fake_belief"/>
</node>
<node pkg="next_best_sense" type="pure_navigation_two_readers" name="nbs" output="screen" args="$(arg pure_args)" if="$(arg two_readers)">
<param name="move_base_costmap_topic_name" value="/map"/>
<param name="move_base_costmap_updates_topic_name" value="/move_base/global_costmap/costmap_updates"/>
<param name="belief_map_srv_name_left" value="/thorvald_left/rfid_grid_map_node/get_rfid_belief_left"/>
<param name="belief_map_srv_name_right" value="/thorvald_right/rfid_grid_map_node/get_rfid_belief_right"/>
<param name="fake_belief_map_srv_name_left" value="/thorvald_left/rfid_grid_map_node/get_rfid_fake_belief_left"/>
<param name="fake_belief_map_srv_name_right" value="/thorvald_right/rfid_grid_map_node/get_rfid_fake_belief_right"/>
</node>
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent