System Racetrack design
0 views
Skip to first unread message
Guy
May 30, 2008, 10:51:27 PM5/30/08
to Team Starfox
Hi Jonathan,
I'm moving this discussion to our group so that any newcomers can
review the history.
> The "System: Racetrack" plans look much too ambitious for just a propeller.
It's definitely too much for the Propeller. In fact this morning I
very quickly entered that diagram from a sketch that I had done on
April 17 when I was discussing the idea with my son. He's the one
that came up with the names for everything. I'm planning to start
working through things in increasing detail and some things are
definitely not obvious at this point.
> How much do you think it can handle?
I think that the Propeller can handle all of the green components. We
need to figure out how we want to load a "mission". I've seen
examples of interfacing the Propeller to SD cards of up to 1GB so
that's an enticing option.
> Where is the Wifi interfaced?
At this point haven't I really investigated the options are but I've
seen various add-on and daughter boards and I expect that we can find
something to fit our needs. What do you figure our needs are?
I would like to start identifying requirements from the mission
planner on down. I have a rough list of mission elements that I'm
going to enter and then start thinking through them.
Guy
I'm moving this discussion to our group so that any newcomers can
review the history.
> The "System: Racetrack" plans look much too ambitious for just a propeller.
It's definitely too much for the Propeller. In fact this morning I
very quickly entered that diagram from a sketch that I had done on
April 17 when I was discussing the idea with my son. He's the one
that came up with the names for everything. I'm planning to start
working through things in increasing detail and some things are
definitely not obvious at this point.
> How much do you think it can handle?
I think that the Propeller can handle all of the green components. We
need to figure out how we want to load a "mission". I've seen
examples of interfacing the Propeller to SD cards of up to 1GB so
that's an enticing option.
> Where is the Wifi interfaced?
At this point haven't I really investigated the options are but I've
seen various add-on and daughter boards and I expect that we can find
something to fit our needs. What do you figure our needs are?
I would like to start identifying requirements from the mission
planner on down. I have a rough list of mission elements that I'm
going to enter and then start thinking through them.
Guy
Guy
May 31, 2008, 9:51:13 AM5/31/08
to Team Starfox
Jonathan,
I've just remembered that at the VanLUG meeting you were talking about
running the high-level control remotely on a laptop. My first choice
would be to make the robot fully self-contained but that decision is
subject to the weight that it can carry which limits the batteries
that it can carry which are needed to run a lot of processing power.
So there's 2 broad design options:
1) Keep it self-contained and limit the amount of processing that we
can do, or
2) Develop using a wireless communications system to allow control
functionality to be run remotely.
For 1) I would focus on using small processors such as the Propeller
or the Arm in the Gumstik. Of course eventually we would outgrow what
we can do on board an R/C model so then we'd either need to migrate to
2) or else use a larger vehicle which is what UBC is doing with
Snowflake.
For 2) the biggest challenge is dealing with the amount of data and
the possibility of communication interruptions. For video I can't see
running a raw video stream over WiFi so I'd either want a lot of
preprocessing or else use a camera with a TV channel transmitter.
Then I'd be interested in using a mesh network and possibly use the
router that http://vonic.ca/ is distributing.
Of course some competitions such as Robo-Magellan preclude any
wireless connections so I'd like to keep 1) as an available
configuration option.
Guy
I've just remembered that at the VanLUG meeting you were talking about
running the high-level control remotely on a laptop. My first choice
would be to make the robot fully self-contained but that decision is
subject to the weight that it can carry which limits the batteries
that it can carry which are needed to run a lot of processing power.
So there's 2 broad design options:
1) Keep it self-contained and limit the amount of processing that we
can do, or
2) Develop using a wireless communications system to allow control
functionality to be run remotely.
For 1) I would focus on using small processors such as the Propeller
or the Arm in the Gumstik. Of course eventually we would outgrow what
we can do on board an R/C model so then we'd either need to migrate to
2) or else use a larger vehicle which is what UBC is doing with
Snowflake.
For 2) the biggest challenge is dealing with the amount of data and
the possibility of communication interruptions. For video I can't see
running a raw video stream over WiFi so I'd either want a lot of
preprocessing or else use a camera with a TV channel transmitter.
Then I'd be interested in using a mesh network and possibly use the
router that http://vonic.ca/ is distributing.
Of course some competitions such as Robo-Magellan preclude any
wireless connections so I'd like to keep 1) as an available
configuration option.
Guy
Guy
Jun 1, 2008, 2:36:59 AM6/1/08
to Team Starfox
I just finished a meeting with Peter and he's interested in addressing
the localization problem. He's interested in building a robotic lawn
mower and wants it to know it's location on the yard very precisely so
he's going to see if he can give that capability to Rawhound. This is
far more accuracy than we need for most tasks but it would allow
Rawhound to do things like mark lines (with a line marking attachment
of some sort). It's a raw idea at this point but I hope that Peter
can use his engineering physics education to pull it off.
-- Guy
the localization problem. He's interested in building a robotic lawn
mower and wants it to know it's location on the yard very precisely so
he's going to see if he can give that capability to Rawhound. This is
far more accuracy than we need for most tasks but it would allow
Rawhound to do things like mark lines (with a line marking attachment
of some sort). It's a raw idea at this point but I hope that Peter
can use his engineering physics education to pull it off.
-- Guy
Jonathan Hanna
Jun 2, 2008, 3:04:44 PM6/2/08
to Team Starfox
I think an interesting system can be made with only on-board control. Whether
real localization can be included I don't know.
With the Propeller we get a max of 160 mips for about 80 mA with 32K ROM, 32K RAM, and another 16K RAM per cog.
(for comparison, a 400Mhz (and about 400 Mips and 64MB) Gumstix is about 250 mA)
Some tasks are:
Poll Bumpers (2)
Poll Hall Sensors (Odometer) (4)
Poll Sonar
Poll Infrared
Poll Compass
Read GPS?
Drive Wheel motor
Drive Steering
Read sound?
Process Sound
Read Video
Process Video (blob)
Process Video (fancier image processing)
Drive Speaker (Emit sounds)
SLAM
Presumably, several sensors can be handled by one cog?
I don't know how easy the programming environment is on the propeller,
but the SPIN language looks a bit limited.
How much battery is there and what is the draw of the motors?
Reply all
Reply to author
Forward
0 new messages