Introducing Computer Vision on a robot requires an upgrade of its computational resources. As soon UDOO Team announced its amazing UDOO x86 board I was enthusiastic imagine which capabilities I can add to Geduino using this powerful board.
As result I’m working to replace the UDOO Quad used by Geduino with a cluster of UDOO x86 with the goal to use one of it for navigation and the second for perception.
This is the result:
The boards are connected by ethernet: the lower board is equipped with the Intel WiFi board (available in UDOO shop) used as access point where external PCs or tablets can connect directly without needs of additional router (some IP Table rules ensure they can access both boards).
Both boards runs Ubuntu Desktop 16.04 with 4.8.0 Linux kernel (be careful: Intel WiFi board is not supported by 4.4.0 kernel!). The desktop manager, network manager, cups and other tipical desktop service was disabled by default but they can still started manually in case they are needed (for example to run the Arduino IDE to upload sketch into Arduino board).
The Geduino Shield has been redesigned in order to fit the new layout and make it compatible with 1.8V logic used by UDOO x86 Braswell side (see my previous post).
Also the Power Board was redesigned in order to provide more current needed by the additional boards.
This is Geduino equipped with the UDOO x86 cluster:
Of course ROS was installed on both boards: it is just designed for distributed hardware so it was quite easy distribuite its nodes across the two boards. As result you can see the RViz output of all Geduino’s sensors (Ping))), RPLidar, Inter RealSense R200).
I’m now working to reactivate the SLAM and Navigation Stack and to perform some tests in order to analyse real workload for the system (CPU usage, CPU temperature, memory consumption, network traffic etc): anyway, at current status, I can absolutely say the the promise of a powerful board was perfectly fulfilled by UDOO Team!