There has always been a human babysitting a cocktail robot, usually by shaking the cocktails for the machines. And I have always wondered, why this is such a difficult task and over the years I have crystallized the following problem:
1) Actuator: Electrical motors are generally too weak or too slow to properly shake a shaker.
2) Shaker: Opening and closing tightly the shaker is even for some humans pretty complicated task, even screwing a jar requires a lot of expensive actuators and a lot of degrees of freedom at which point, coordinating the movement is getting complicating requiring hundreds of hours to implement.
3) Mixing: Usually done with time controlled peristaltic pumps, usually pretty slow thus not really enjoyable.
4) Serving: Getting the cocktail out of the bot can be sometimes tricky and this is what makes it usually look like a huge coffee dispenser, which I hated and the reason is, putting liquid from one cup to another has the same problem as 2).
5) Cleaning: yeah
-> I have 21 Weeks, so for each step I can use 2 Weeks roughtly, I will be busy with other stuff in life so wouldn't be too generous with it.
Week 1: Compressor
Week 16: Testest some neumatic parts for the compressor, but cannot decide for a final design and thus postponed the deadline for one year. Should probably make a blueprint as I did in my good old days with pencil and paper.
4,4cl gin, want to use quality gin like "The Duke" (external link) 0,74cl triple sec, might be replaced with hops-liquor by "Hopfa" (See external link) 2,2 cl lime juice 4,4cl orange juice 0,5cl grenadine
Total 12.24 cl
Is a usual serving at a bar from my experience. It is actually measured in volume but for my project I will need this recipe in grams, which we will do in the future with a syringe and a precision scale. I want to measure the bottles with the liquids constantly to be able to measure out the liquids and not running on empty, which can be a little bit fatal, because I want to pressurize the bottles (not too much) like a kegerator, in that way, I don't need the ugly and slow peristaltic motors.
I have once made a simple robot arm with just a dc motor and at the pivot point a potentiometer. This robot will get ice from a rotatable carrousel with dispensable cups (hopefully paper, since it has a flat bottom and more environmentally friendly) and everything will be controlled via mqtt on a raspberry pi and a 12v simple router (not running openWRT....yeah kinda defeats it's purpose). I chose mqtt because it is flexible and asynchronous and i love ESP8622, and I want to expland my knowledge with ESP32 this time, I hope it does not escalate too much. This approach keeps the option open to have a ESP or even nodeRed server on the raspberry to control the whole thing, which does not involve tiresome debugging. It decouples the application (making cocktail) from the firmware (controlling motors) which is a good practice, since my things are usually hardcoded, which is not a good practice.
I specifically chose one cocktail, because I wanted to focus on the reduction of human interaction, because I don't like babysitting robots. Makes me feel like a bad cog in a machine. I have been at roboexotica many times also as exhibitor, and you have to explain your robot everytime to everyone to sell your cocktail to finance back a little bit, because making robots is usually not cheap and I want to refund my travelling costs.
The idea for the cocktail shaker took me two years to come up: There is a stainless steel container with acrylic lid which is hinged. This hinge will be the middle point for some aluminum profile piece which is connected with a pneumatic cylinder. My hope is, that the pressure of the pneumatic cylinder is great enough to close it airtight but easy to open with the pneumatic cylinder. On the lid, there is a hole for stainless steel quick push. There will be only one hose and it will be connected to a manifold which is connected to the bottle of spirits and water. The bottles has air inlets with seperate electromagnetic valves, and i hope that the low pressure will be enough otherwise i have to put seperate flow regulators on each valve afterwards. Need to test. Currently I am not sure, whether i should open the valves seperately or at the same time.
This shaker is mounted on the right arm which is pneumatic. It should be removable for cleaning. There will be a load cell on lower arm, not taking the full load like the openDog from Xrobots but mounted in parallel. My hope is, that through the maximal weight, an appropriate acceleration can be calibrated or maybe just a threshold. but definitely a interesting data to look at. Because shaking requires timing, which cannot be hardcoded, otherwise it won't shake well. Since pneumatics cannot be really well position controlled (possible but requires precise pressure difference controlling with some absolute measuring point) i would mount several hall sensors to also trigger the shaking motion. I am currently thinking whether I should use a AS5600 magnetic encoder or something to be able to control the motion more precisely.
It is my vice and virtue to pursuit beauty in motion and my ideal shaking technique is shaking up and down with one arm. You can find it in the link collection under "Shaking technique". I could have made...