Fontys Adaptive Robotics Minor, Cooperative Robotics Project: Progress with multi-robot exploration

Currently we are working on the multi-robot exploration and mapping. This project is directly connected with the topic of robot vision. Initially we created a simulation of our system in stage, where we put three robots that were using slam-gmapping and auto exploration. The robots send 2d map data between each other using a simulated ad-hoc network. With the help with a ROS package called map-merger the robots could find common similarities between maps and connect them into one global map. Bellow you can see the Stage simulation, and the map merging result

In order to make this possible, we had to write our own launch files with specific parameters.
After the successful simulation we moved on to a real world platform, which is Rosbee. It’s a very simple ROS platform with two wheels with wheel encoders. Our first challenge was to have the robots up and running. We had to repair minor stuff in the firmware, such as the direction of rotation when turning left/right. We also added small hardware accessories that make the work with Rosbee easier, such as a power switch, a power connector for hokuyo laser and a usb power adaptor for connecting the rpi.

After we had three fully working Rosbee robots we moved on to bringing our simulation to reality. Our first impression was that we would be able to run all the ROS packages on a Raspberry pi3, however that was not the case. A Rpi3 is able of running the movement and sensor reading, however it’s not able of running map merging algorithms which requires advanced image processing. As a solution to this problem, we decided to run the rosmaster and the map merging packages on a remote PC, where we can also visualize the data. On the raspberry pi we have only laser node, navigation stack and gmaping.
Bellow you can see the rqt graph of an individual robot and laptop:

If we subscribe with rviz to the map_merger node, we can see the global map, and if possible the position of other robots in the ad-hoc range. Below there is the result of merging maps from two lasers on top of each other.