Here is an assumptions based customer persona brainstorm and an assumptions based current state user journey map for the group “creatively interested professionals”. We also added in some places that our coffee delivery robot KRYTN could help.
We learnt a few things from this:
1, The barista talks a lot and using voice might interrupt their chats. So let’s scratch the voice control feature of KRYTN.
2, The robot can likely improve the lowest points in the coffee service offering - waiting for coffee. The customer observing others getting served by a robot should be quite engaging and fun and hopefully lead to anticipation for them.
3, When Dharmesh Patel is operating solo he can lose track of what he is doing when delivering coffee to customers. So KRYTN can potentially improve his workflow and ability to make more coffee.
Note these are all assumptions until we test them, but at least they can guide our design decisions.
Is the name KRYTN inspired by British sitcom Red Dwarf? If you you are OK in my book! I have seen all episodes through Red Dwarf 7 or maybe 8 and Kryton ROCKS! Brilliantly irreverent series!
I realize your last name starts with V but I will ask of you know about the sort of silly movie called ‘Short Circuit’ and if so, my hat is off to you for your most excellent Avitar choice on this board.
Haha yes! KRYTN is named after Red Dwarfs Kryten its abit of a silly project so it needed to be named after quite a silly robot
And wow, yes I really enjoyed Short Circuit when I was younger, actually I tried naming my github johnny5 for that reason (but had to settle for johnny555). It never occured to me that johnnyv could mean the same anyway! Thanks @mjstn2011!
When designing for humans, how do you know if your concept will work? Easy, test it on yourself! (Assuming you are human right?)
In this video Dharmesh Patel , Hilary Goh and myself are doing an after hours service rehearsal with KRYTN in GAMECITY coffee shop. The idea is to test the prototype and see if there is any awkwardness on the human interaction for the concept. Besides my inability to drink properly and KRYTN’s poorly tuned navigation stack, this went well. KRYTN’s height is only a little low but for someone sitting on the gamecity cubes it’s ok. Importantly, the robot didn’t freak out that there was a coffee container directly behind it the whole time.
Next steps are tuning the navigation system and building an interface for the barista to use.
For people watching closely you also get a glimpse of Hilary Goh’s spipy hexapod as well
I take it this is using Lidar for the navigation?
If you slow it down it may make less ‘rude’ course corrections AND maybe not spill andy of the precious life giving java!
Are you using a path planner or is this your own ‘nav’ sofware?
Hi @mjstn2011, its using both fiducial_slam using ceilings and lidar.
However, the fiducial_slam is quite jittery, I think because the roof in Game City is quite high (it works pretty good in my house though). So the robot relys strongly on the lidar map, but still often gets lost as the Game City is just one big rectangle.
I tried the nav tuning parameters from the below repo, and by default it does move much slower but did work much better in later testing:
I’m hoping my ultimate solution will include both fiducials and lidar though. I’m waiting for a raspicam HD to turn up so I can trial with a better camera.
Has anyone used a raspicam HD with Magni? How have you installed it physically?
I have a Raspicam HD that I use with an Arducam 6mm lens. That lens is simpler and shorter than the HQ lens and holds focus easily. Another member of our team is working with the stock Raspicam HQ lens. We are evaluating it.
As for how we attach them, I have made a 3D printed adapter that allows screwing down the adapter to our existing upward or forward mounts and then screw in the Raspicam HQ onto that plastic piece using M2.5 wood screws of 5mm length. It works well.
Contact me if you want more detail on the mounting adapter. Use this forum because my email is giving me grief this week.
If you use the Feb 2020 image the support for the camera is present on that image.
I have a quick n dirty camera cal for the 6mm arducam lens on the raspicam HQ but I am wanting to re-do that calibration as it is not good enough. We are trying to stick to lenses that do not require ‘fish eye’ mode for the calibration. 6mm seems a good compromise and can ‘see’ the full chip it seems and yields a 65deg HFOV. I tried the Arducam 4mm lens for raspicam but you get dark corners but 95 deg HFOV. You may be able to ‘crop’ the images with some option and get 80deg HFOV with no dark corners but I have not tried that yet.
Are you located within the USPS CONUS spaces? If so I can just mail you one on the house. (gotta fire up the ol 3D printer and make a couple copies first).
Don’ t post any address here, just let me know if it’s easy for me to send and not international or other complex thing.
Thanks @mjstn2011 ! thats very generous. Unfortunately I’m not in CONUS (actually based in Perth, Australia), but I’ll send you an email anyway and see if we can work something out
Hey @mjstn2011 I got the part printed but I think the top tray needs to be removed for it to fit. There really isn’t enough clearance as you can see from the photo
I cut my top tray with a hack saw. I am also using a far shorter lens then the stock Raspicam HQ. Look into the Arducam 6mm lens was my suggestion, NOT the stock RaspiCam really tall lens. I was trying to convince persons in our organization for support of these things but it is a slow process and REQUIRES users like you to say ‘yes we need this camera’. There is also one more ‘just about ok but not ok’ clearance issue. The adapter plate 3D print design I sent you has the RaspiCamHQ pcb basically rubbing the metal standoff. So I think it needs a couple mm more spacing but it is sort of ok. It hits the upper plate standoff. Pictures would be better here but in short we need to work out proper clearances. I know that. That is why it is not ‘officially’ approved. I was thinking of making another version of my 3D printed adapter to move the camera out in +x direction by 3 or 4mm which would have some impact on or robot model but would make the fit cleaner. The raspicamHq is just a lot larger PCB.