Fischertechnik Iot, virtual reality and Machine Learning works: National University of Colombia

Alles rund um TX(T) und RoboPro, mit ft-Hard- und Software
Computing using original ft hard- and software
Forumsregeln
Bitte beachte die Forumsregeln!
Antworten
lorquenix
Beiträge: 1
Registriert: 23 Okt 2019, 21:25

Fischertechnik Iot, virtual reality and Machine Learning works: National University of Colombia

Beitrag von lorquenix » 23 Okt 2019, 22:54

Hi!
My name is Carlos Enrique Mosquera Trujillo, I am fourth semester student from electronic Engineering of the National University of Colombia campus Manizales. I am in an research group called STEM and we have worked in fischertechnik things almost a year. My work on the group is develop techniques of Iot connections like MQTT in the TXT Controller. First I try to do that in te Robopro`s software but was impossible, later I discovered python´s library FTROBOPY and implemented it with Paho-Mqtt Library, I was able to do little things like control the Iot Module, Vacuum gripper robot and take statistics of the sorting line. Here is a little video of that.

https://www.youtube.com/watch?v=IptTJqq9D5I

This technique has a problem, you can only introduce the Iot capabilities with a computer or a raspberry, but you can not run directly in the TXT controller because the Mqtt library is not compatible with the ARM architecture, I have try a lot to install it with not success. After a long time doing research in the SSH terminal and the c++ programming (I sincerely believe that it is poorly documented), we develop a simple library which can run IOT programs in any TXT controller directly. How it works?, well i have found how to redirect the connection to fischertechnik cloud of the Smart Home program, and discovered that in the TXT is installed Mosquitto which implements the MQTT protocol in the terminal. In python you can put commands in the terminal and log them, this is the trick of the library. To run the python script, you can do a c++ program which you can start without terminal.

Well, other works in the stem group are module detection app realized in tensorflow, we take almost 2000 photos and make a convolutional network:

Bild

The most important project that we are development is a virtual reality software to do trainings in industrial works:

https://www.youtube.com/watch?v=y9ChVe9BLyE
https://www.youtube.com/watch?v=VZNX8kMg8tA
Bild
If you want to contact us, fell free to write in this forum or in this email:

cmosquerat@unal.edu.co
Zuletzt geändert von lorquenix am 23 Okt 2019, 23:40, insgesamt 1-mal geändert.

Benutzeravatar
PHabermehl
Beiträge: 2429
Registriert: 20 Dez 2014, 22:59
Wohnort: Bad Hersfeld

Re: Fischertechnik Iot, virtual reality and Machine Learning works: National University of Colombia

Beitrag von PHabermehl » 23 Okt 2019, 23:30

Dear Carlos,

This is quite impressive work! Seeing a ft factory in VR is abolutely fascinating.

Welcome to the forums and - please keep us up to date!

Kind regards
Peter
https://www.MINTronics.de -- der ftDuino & TX-Pi Shop!

viele Grüße
Peter

vleeuwen
Beiträge: 1557
Registriert: 31 Okt 2010, 22:23
Wohnort: Enschede (NL)
Kontaktdaten:

Re: Fischertechnik Iot, virtual reality and Machine Learning works: National University of Colombia

Beitrag von vleeuwen » 07 Nov 2019, 15:47

Since Firmware 4.5.1 Mosquitto MQTT libraries are available on the TXT; so they are running on this ARM-machine.
On the MQTT web site are examples available.

The fischertechnik TXT-Trainings_factory software gives a good example of how these libraries can be used. The MQTT client can also be used in another application.

The fischertechnik txt_demo_c_download shows the basic of local C/C++ programming (IoT).

The FtRoboPy is localy slower because it is their also using the (Berkeley) network or ip-socket connection with the TA.

In the TXT-Trainings Factory, the MQTT-server/bridge is running on a 5th TXT, but you can locate this server also on a other system (PC, laptop) in the locale area network. The Mosquitto libraries for Windows can also be found on the Mosquitto web site.
Zuletzt geändert von vleeuwen am 07 Nov 2019, 18:51, insgesamt 4-mal geändert.

Benutzeravatar
EstherM
Beiträge: 1466
Registriert: 11 Dez 2011, 21:24

Re: Fischertechnik Iot, virtual reality and Machine Learning works: National University of Colombia

Beitrag von EstherM » 07 Nov 2019, 15:52

Wenn Du diesen thread gerne auf Deutsch lesen möchtest, gib bitte dem Moderatoren-Team bescheid.
(If you would like to read this thread in German, please contact a member of the moderator team).
Gruß
Esther

Benutzeravatar
MasterOfGizmo
Beiträge: 2720
Registriert: 30 Nov 2014, 07:44

Re: Fischertechnik Iot, virtual reality and Machine Learning works: National University of Colombia

Beitrag von MasterOfGizmo » 07 Nov 2019, 16:49

vleeuwen hat geschrieben:
07 Nov 2019, 15:47
The FtRoboPy is localy slower because it is their also using the winsocket connection with the TA.
This is wrong. Winsocket is a windows API (hence the name) and does not exist on the TXT. You are probably referring to the Berkeley socket interface. This interface is commonly used by servers for pretty much everything that's network related. And this is not per se slow especially not when being used on the same device (localhost) and for such small and slow things like the RoboPro socket interface (the thing you call TA). It may be slower than a shared memory solution but the linux kernel is pretty good in figuring out that you are just sending data forth and back on the same machine.

Das stimmt so nicht. Winsocket ist eine Windows API (daher der Name) und existiert auf dem TXT nicht. Du meinst wahrscheinlich die Berkeley-Socket-Schnittsstelle. Diese Schnittstelle wird üblicherweise auf Servern für praktisch alles verwendet, was mit Netzwerk zu tun hat. Und das ist nicht per se langsam, speziell dann nicht, wenn die Kommunikation auf dem gleichen Gerät abläuft (localhost) and uch nicht, wenn es um so kleine und langsame Dinge wie die RoboPro-Schnittstelle geht (das, was Du TA nennst). Es mag langsamer sein als eine sogenannte Shared-Memory-Lösung aber der Linux-Kernel ist ziemlich gut darin, festzustellen, dass Daten nur innerhalb des gleichen Gerätes ausgetauscht werden.
Arduino für fischertechnik: ftDuino http://ftduino.de, ftDuino32 http://ftduino.de/32

vleeuwen
Beiträge: 1557
Registriert: 31 Okt 2010, 22:23
Wohnort: Enschede (NL)
Kontaktdaten:

Re: Fischertechnik Iot, virtual reality and Machine Learning works: National University of Colombia

Beitrag von vleeuwen » 07 Nov 2019, 17:09

Sorry: a small type error.
On the TXT it is an implementation of (Berkeley) Network-socket.
The Winsocket is the implementation of a Berkeley socket in the MS-Windows libraries. The basics are the same.
The name ip-socket is often use to express that is an internet (ip-address) relate use of the socket and it does not indicate the exacte datagram here.

The TA (transfer area) is the name used by fischertechnik for the communication data area. It is well documented in the header file FtShmem.h and in the older documentation for the TX-C (see the documentation section English and German) and Robo-Interface documentation.

However the essence of this thread is the availability of Mosquitto MQTT (client) libraries on the TXT.
And were to find examples.
Zuletzt geändert von vleeuwen am 07 Nov 2019, 22:17, insgesamt 7-mal geändert.

Benutzeravatar
MasterOfGizmo
Beiträge: 2720
Registriert: 30 Nov 2014, 07:44

Re: Fischertechnik Iot, virtual reality and Machine Learning works: National University of Colombia

Beitrag von MasterOfGizmo » 07 Nov 2019, 17:19

Er schreibt, dass er sich vertippt hat und dass auf dem TXT Berkeley IP -Sockets genutzt werden und Winsockets eine Bibliothek in Windows ist.

Stimmt zwar nicht so ganz und der TXT nutzt an der Stelle TCP-Sockets und keine IP-Sockets,. Aber seine Antwort wiederholt so ungefähr was ich auch schon geschrieben hatte.

This is not entirely correct as the TXT in this case uses TCP sockets and not IP sockets. But his reply repeated roughly what i already wrote previously.
Arduino für fischertechnik: ftDuino http://ftduino.de, ftDuino32 http://ftduino.de/32

Antworten