20 Sep 2003

Sat, 20 Sep 2003 19:22:00 GMT

Almost a year now from my first an last diary entry...

The "coupe de France de robotique" (French robotic cup) took place in may, with more than 189 teams registered. Botzilla was among the 144 qualified, and after the fourth day and we ranked 49th. Some (non-technical) pictures can be seen here. The official site for the 2003 competition is here (in French).

It still hurt to say it, but though this was a success for the team this was a failure for me: we had to completely change our strategy 1 week before the competition, relying entirely on a PIC16F877 which original purpose was only to manage the propulsion wheels. The plan was to use a 68332 board running a real-time executive for the high level code, but though everything was OK when running in RAM, I failed to make in run in flash, thus preventing the robot to run autonomously. I obviously grossly underestimated the difficulty in making my BSP running from flash, starting to work about it something like 2 weeks before the competition after almost a year of work on it (but work in RAM!). I must have messed up with the chip select settings... Stupid me. Well, anyhow a time came when we had to make a decision and we backed-up on a state machine on the PIC.

My 68332 board stood in its box since we came back from the competition more than 4 month ago until today. Sigh.

On the positive side, as usual, much was learned. You can still do a lot with an 8 bit thingy like a PIC, though if it was to be done again I would vote for an Atmel AVR (the lack of real stack on Microchip is a very real limitation). On the 68332, I played with the TPU, a kind of coprocessor dedicated to time-related function like PWM and such. The trick is that you can re-program the microcode of this TPU in order to use or create the functions that fit your application need. I used some microcode from Arno Morbach providing I2C capabilities to the TPU. I assembled this on Linux using TAS, Zoltán Kócsi's GPL TPU assembler. Then I glued all this with RTEMS and the TPU lib from the motorobots project. And though quite beta, all this actually worked. The TPU-I2C function was previously know to work with serial EEPROM, I can tell that I had it talk with some PCF8574.

I didn't do any kind of serious (non-professional) work this summer. Robotic hacking is a winter sport.

BTW, if by looking at botzilla's picture you wonder how we managed to get a round shape out of wood: this is a special plywood, called "souplex" (in French at least). This is used for curved furniture, like bars and such. You can find some at a professional wood dealer.

23 Oct 2002

Wed, 23 Oct 2002 14:18:05 GMT

Here is the longer story: how it started

I'm a 0x24 years old French geek. I'm a software engineer in the embedded field with a background in industrial automation and electronics among other things. A GNU/Linux user since 97 I think, heavily since 99, mainly since 2000.

During the summer of 2001, I looked for an embedded target to play with. My goals were:

to explore a classic but non-Intel architecture (my job involve Intel's CPU at a very low, architecture dependant level),
to experiment GNU and OSS tools in an embedded development context, after a few interesting experiments with Linux kernel driver and RTLinux modules.

I chose a German made Motorola 68K board, the NF300. Designed in an academic setting for pedagogical purposes, it offered the ubiquitous and GNU- supported 68K/CPU32 architecture with a good level of hardware documentation for a reasonable price. It also offers quite a few peripherals to play with without bothering adding custom-made hardware.

I started scanning through the whole NF300 and Motorola documentation, played first with the delivered DOS tools, then build a cross-GCC tool chain from scratch on Linux and its BDM cross-debugger and... And then time passed but I still didn't have a realistic but challenging project to actually do something out of it.

In the meantime my brother Gregoire became interested and this little board with the flashing LED sitting across my monitor. Working as a research engineer for the French National Geographic Institute, he's a geek too but in the scientific computing field. Dominique, a friend since 0x0f years was quite curious too. Dominique comes from electronic and happens to be good at mechanics too (though he would deny it) and yes, he can be described as a geek too.

The three of us shared common interests and curiosity for each other's knowledge, and started to think of building something together. In the winter of 2001, we saw a small robotic competition. That was it. A robot. We could do one. We (are supposed to) have the skills.

We wanted to have a "bill of contract for it". Requirements, deadline. We wanted rules, to have a clear goal and motivation. We needed a competition to provide these rules.

I started to look for robot competitions in France or near Europe, and quickly realized (without much surprised) that there weren't much, and worse they were all for students. We are not students. Sorry, not much we can do about that. The more interesting competition by far was the two competitions organized by the ANSTJ (Association Nationale Sciences et Techniques Jeunesse = Youth Science and Technical non-profit Organization): the Coupe de France de Robotique (in French) and the Eurobot . These are in fact one contest in two, a national and a European, sharing the same rules. I contacted the organizing ANSTJ and was informed that, though not automatically guaranteed, non-students teams could possibly compete, depending among other things on their input to the general and technical mood of the thing.

The project was launched. I chose "botzilla" as its name (though there are tons of *zillas out there that's a fun and easy to remember name).

To be continued.

robots.net blog for jbm

20 Sep 2003

23 Oct 2002