Re:Building a Robot Dog (with an airsoft gun)
by eschluntz on Wednesday October 29, 2025 @01:15PM
(#45749936)
Blog post about my mechanical design, robot math, stress testing, and competition for "Mech Warfare", a small airsoft-only battlebots competition this summer
Re:Building a Robot Dog (with an airsoft gun)
by gertlex on Wednesday October 29, 2025 @03:28PM
(#45751815)
I can't quote it offhand, but it's stuff like:
- Don't stick your camera on an arm and poke it around corners to see the opponent
- No detachable parts? It's more nuanced than this. Not leaving "mines" I think is a case of this.
Basically there's a few goals to the rules (this has been an on and off competition for about 15 years now): keep the spirit of "pilots in giant mech robots" (loosely 90s era stuff), and homogenize the competition slightly for competitiveness.
Melee is still allowed! If you walk into the other robot and trigger one of their 4 score panels, that still counts.
It's a hard competition to prep for. Numerous bespoke HW subsystems, software, budgets that grow quickly with the size/weight of your bot... and being reliable.
Re:Building a Robot Dog (with an airsoft gun)
by djmips on Wednesday October 29, 2025 @03:17PM
(#45751654)
Excellent write-up. This is one of those side projects that has been on my list since forever and this seems about the level of accomplishment and tech that is within reach. Good effort! I'm looking forward to further refinements.
Re:Building a Robot Dog (with an airsoft gun)
by gertlex on Wednesday October 29, 2025 @06:32PM
(#45753956)
Having had a chance to read more of the blog now, maybe the most pleasing part is your similar discovery of, "I'm writing my own kinematics, and generating plots of it is the fastest validation of that code". I did similar matplotlib plots of my quad (spider) leg movements years ago. Definitely need to do more of that when I finally revisit the software on my mech from 2019 with AI help :).
And more generally, yeah, having the AI manage some of those plotting needs will be fantastic in other work too; I enjoy figuring out what to plot, and don't mind wrangling and hand-refactoring said scripts... but it's so time consuming historically, and there's always the itch to urgency to get back to the "main code". (at least I learned to gradually convince peers that often plot scripts should be commitable without being polished; they're a form of documentation sometimes!)
- guy whose back yard the scrimmages took place in
- Don't stick your camera on an arm and poke it around corners to see the opponent
- No detachable parts? It's more nuanced than this. Not leaving "mines" I think is a case of this.
Basically there's a few goals to the rules (this has been an on and off competition for about 15 years now): keep the spirit of "pilots in giant mech robots" (loosely 90s era stuff), and homogenize the competition slightly for competitiveness.
Melee is still allowed! If you walk into the other robot and trigger one of their 4 score panels, that still counts.
It's a hard competition to prep for. Numerous bespoke HW subsystems, software, budgets that grow quickly with the size/weight of your bot... and being reliable.
* Z3l3nsky slaps eschluntz around a bit with a large trout
<Z3l3nsky> @eschluntz pm
That happened.
And more generally, yeah, having the AI manage some of those plotting needs will be fantastic in other work too; I enjoy figuring out what to plot, and don't mind wrangling and hand-refactoring said scripts... but it's so time consuming historically, and there's always the itch to urgency to get back to the "main code". (at least I learned to gradually convince peers that often plot scripts should be commitable without being polished; they're a form of documentation sometimes!)
https://web.archive.org/web/20251030055726/https://erikschlu...