Design notes on each version of PixelBot. Select a tab to learn more.
LAST UPDATED: OCTOBER, 2022
TO DO LIST:
- Finish the FollowBot Command Generator tool (currently in beta). Should be mobile friendly and sort out a few bugs.
- Build new iteration of launcher:
- New motors and suspension system
- New loading servo and jam protection
- Draw/animation Editor for eyes tool.
- Animation editor for radar tool.
Finished the body, and built 2 new iterations of the launcher to help increase DC motor lifespan. The latest launcher iteration has performed the best, but still needs work. Biggest addition is I built an automatic reloader system so I no longer have to reload (unless viewers shoot over 1000 shots). It uses a gravity-fed hopper and air exchange system to push the ping pong balls to the bot. It is complex enough that I will do a separate post on the system.
I finally ditched the BB throwing arm for a less-messy prosthetic arm.
Latest left arm. When he raises his arm the specially-designed wrist does a salute action.
Ditched the cross-bow design for a triad design that uses a top idler to maintain pressure against the ball during launch. This allowed me to lower the driving voltage on the motors extending the life a bit.
Split design so I can swap the front end out for a different design in the future.
Latest launcher being used. Back to cross-bow configuration, but Inverted the motors so I do not run them CCW. This helped a lot.
Ball feed - magazine coupling. It has a slip connection so magazine can rotate in the coupling.
The hopper. The top drum holds 1000 balls and has an agitator to force balls down the feed tube where a blower takes over to push balls to FollowBot.
- The launcher still chews up motors over time. It appears that it is less brush wear and more bearings now. I purchased fancy motors from Out of Darts, and sketched up a spring suspension system for the motors that may give the launcher long life. If that fails I'll move to brushless motors.
- Occassionally, I burn out the servo that pushes the ball into the launcher. Though I have code that protects against this it seems to be an astable edge condition on startup or shutdown of the node server. Consider adding spring action in the loader arm so it doesn't lock up. Also, the tactic servos are no longer available. Need to redesign mount for a more common size (EG: micro servo) and one that is cheaper since only requires low torque.
For the next iteration, hacking a PTZ camera system would have been easier (like Predator shoulder-mounted laser), but I wanted the bot to be large and highly animated for visual entertainment purposes. I designed a 30-shot magazine for the bot. It had to flex since he can pivot 180o. I came up with using 3D printed, modular, spring-like sections to form the magazine.
The body was built, LED eyes and the radar features added. Stuck with the same cross-bow POC launcher.
Even though I increased the magazine capacity, I was still spending most of my time reloading.
I kept the BB throwing arm for this iteration.
- The body was initially built from 1/4" MDF. It was too heavy for stock, high-torque servos to direct drive the waist. Options were to add a gearbox, buy an expensive servo, or get the weight lower. I chose the latter to keep complexity and cost low. I rebuilt it from 1/4" Birch plywood and dropped the wieght by 60%. Works nicely.
- The waist bearing system is pulled from an old, high-end security cam system that had a fancy slip-ring to pass signals/power through. For final I will need to redesign this for an off-the-shelf bearing.
- I initially drove the LEDs with the same Arduino running the bot. When people animated the eyes it would cause shuddering and glitches in the bot performance. The data rate was too much for the single Arduino. I added a secong one to solve the issue for now. Eventually, the entire system will use a single-board computer like the Raspberrry Pi.
- All body panels are removeable for future servicing and modification.
I built a proof of concept together using a cardboard box and a few other laser cut cardboard parts. As the code evolved more complexity was added.
The very first iteration sat on my desk and threw airsoft BB's when someone "Followed" my Twitch channel.
Next I added a rudamentary ping pong ball launcher. It used a single-wheel launcher system. Underpowered but functional.
Built a powerful cross-bow-style 3D printed launcher with a 6-shot capacity and added pan and tilt for targeting. This is where the chaos began.
The POC2 launcher. Still mostly hacked together. Ball sensor let's me know when to reload. Built a printed counter-balancing spring to support the weight at rest.
The ball feeder arm was the best part of the design. With 1 servo/action it will control the feed and push the ball into the rotating wheels.
- The cheap brushed hobby motors will burn up over time. I run one of them CCW to maintain symmetry which is not good for the brushes. Need to consider this in the next design.
- The 6-shot magazine is not big enough. I spend most of my time during stream reloading the bot. I need to expand the magazine and eventually consider an automatic reload system.
- I probably will retire the BB throwing action. This works nicely when on my desk and people can see the mess it creates, but once moved to the back of the room it isn't visible.