It might be a good idea to use a clutch system with permanently meshing gears. That would allow you to switch outputs quicker and address multiple outputs at once if they all need to be turned in the same direction
I prototyped such system for a loading/unloading filament mechanism, using small clutchs from copiers. The issue is they went hot when energized, and melted my PLA parts :o/ I may try again with ABS, one day.
I feel like this could be used for a self leveling platform. A table would be the easiest thing to say, you can plop it anywhere and the table will adjust the leg height automatically to become level.
I've been trying to think of a design for a washing machine stand that I can level more easily than adjusting the feet of the machine. This mechanism is better than any of the designs I had in mind.
You could use this to keep soil irrigated for a 3-dimensional stack of plants. Moisture sensors could tell you which plant needs water then this could be used to send the dispenser to the proper xyz position.
Wonderful idea! But the idea of a moisture sensor would be troublesome from the perspective of establishing values. Thus could be bypassed via a little chip that gets programmed with set intervals, or maybe a knob that adjusts the timing by 5% every revolution.
@@garnerday7149 farmers just use drones or quadcopters with cams for multi-spectral images. You don't really need a boots-on-the-ground solution for that other than in remarkable spots. Some trained sw and that kind of data would handle much of what could be automated.
Yes, and also multiple perestaltic pumps, the most awesome water valve of them all XD Anyway... retrofitting existing water valves would save them from the trash! They can be quite demanding in torque!
Robot bartender? Watering a bunch of plants? Might be interesting if you could make the multiplexer circular. That way you’re never more than half the outputs away from your next position.
@@travisash8180 As far as I know FDM 3d prints are rarely watertight out of the printer, so it might be better to use a kitchen funnel from a dollar store or something like that
This can be used in a flip clock. It will rest on the 'seconds' position, but will move along when the 'hour' or 'minute' change. Could even be extended to a calendar or any other kind of counter.
Sailboats often have an assisted motor for pulling anchor chains, hoisting sails and adjusting the sail position, people usually manually change the input to what they need.
It’s awesome that you share these videos, but the even bigger thing is that you provide a literal master class in critical problem-solving, as it relates to identifying all case use scenarios. This video is a great example - you walk us through, step by step, how you address each potential scenario and subsequently beef up your code to resolve the issue in question. This sort of critical logic application is less common than we might collectively hope for. 😂 As an aside, my daughter is 8 and I promised her we would build one of your robotic dogs - looking forward to the challenge!
I recently did a project very similar to this, but with few differences. The main motor was connected to a long lantern gear, mounted to motor shaft. Corresponding star gears were engaged to this lantern as the outputs. These star gears were mounted on a component, which featured solenoid, with mounted gear. When the solenoid was activated, whole mechanism engaged and torque was transfered. Advantage of this approach is that we can have more than one output engaged at the time.
I didn't see it at first, you did a great job presenting every step....then it clicked. Nicely done. Could you add two more axis, a joystick(or integrate like a game controller) and an arm with a cup? Excavator, or those could be the programing inputs for like a pick and place arm, safe cracker, laser guided drill press, etch a sketch with retractable pen....damn the possibilities.
Have your TV or PC screen oriented at optimal position. With the proper sensors it could follow you around the room. This would be awesome in a home gym where one has multiple exercising machines spread throughout the room or while working on a project in the garage/shed.
this would be great for some kind of sequence of events, like a "loading" robot that first grabs a ball/projectile, then swings a arm to a receiver then releases the clamp and then shuts the receiver hatch. after it will then reset all items to their original position
@@brendandor I'm thinking phase 1 a catapult winding back, phase 2 something dispensing or an arm dropping some type of projectile in, and phase 3 release to fire at a predetermined position like the entry way to your room just for fun
If you used a double clutch system on each output, you could have a motor thats turning at a constant speed, and outputs triggered simultaneously in either direction 🙂
Made something like this in college. It had twelve outputs that were vertical screws driving an indicator. The planned use was to indicate the tide for the next twelve hours.
You could turn arround the first worm gear so that there is less distance between the three input gears. That reduces travel time, espacially between the first and the second input. Then you could hook up the most often used output to the worm gear in the middle, and the second most often used output to the first worm gear.
This seems almost identical to a manual gearbox, just with multiple outputs instead of a single output shaft and a moving motor instead of clutches. Rather than moving the motor and using the gears as a clutch, all the gears and motor would stay fixed and meshed, and clutches would engage or disengage each output. This would also have the advantage of using multiple outputs at once rather than being stuck with one at a time. Plus much faster engagement.
With friction clutches it would make a good tracked vehicle drivetrain, or 8x8 wheeled vehicle. Engaging one side or the other to steer, and if using wheels only engage the wheels that are on the ground.
This mechanism remembers me so much of the Prusa MMU, it's basically the same exact concept for multiplexing multiple filament to a single filament extruder.
Mechanical multiplexer is an amazing idea. It is a perfect solution for a robotic arm as each degree of freedom needs high torque and high precision = big and expensive motor. Unused channels would require breaks though. I am looking forward to the future videos on this topic.
There is a lot of knowledge and know how behind making something working as smoothly as shown. It might look easy but it is certainly not. I appreciate the quality of things on this channel.
Makes me wonder if we can benefit from multiplication. 3 data points can control 7 different variables, 3 completely. 4 can control 14, 9 completely. (Math done in head while exhausted at 1:24am) There should be a way to mechanically pull it off. Example, say gear A, B, C, and D. AB, AC, AD, BC, BD, CD, ABC, ABD, and BCD can be controlled pretty accurately while varying from each other by just the 4 inputs. It costs a bit of accuracy, but can increase the amount of control through minimal gears. And accuracy increases as precision does. Hope this makes sense. Really worn out ATM.
James: Builds SLAM-located, AI navigating quadrapedic robot dog with ROS and advanced power electronics. Also James: "So I just used tape and zipties to attach these potentiometers." That's the expertise of a true engineer!
This is used in high school robotics commonly for alternative modes of robot movement. So in the case the robot has to climb there's really no point to also be able to drive at the same time. So it could switch between drive and climb as once lined up for climbing the drive doesn't matter.
Hi James, as a LEGO fan I didn’t have that many motor, so I used this system of a sliding motor to many outputs a lot. I had a good inspiration with Technic set 8082 Multi Control Set. You could build 4 models with this set and one of them was a skid steer forklift vehicle with this system. I used it the control a self designed crane to move the boom, arm, hook and rotate the crane. With 2 motors and 2 up and down joysticks I could move a model with 4 functions. This was absolutely great. Now I have a hole draw full of motor so I don’t use this system any more, because it is very bulky and only works in large models. If you interested in the LEGO set, there are many website that have LEGO instructions to watch online or download in pdf. It is quick a interesting model, because it is very compact and the positioning motor only moves the rack 1 stud per function. So to compact this system you could learn a lot from this model :) You could build a rover that moves forward and back, slide and it turns left and right, slide and the arm moves up and down, slide and it opens or closes the grabber. With rotation sensors you could program it to move a programmed path or with a light sensor make it a line following rover. Make it pick up objects or even give it a fifth function os it can alternate between different attachments. Or rotate the grabber so it can pore a can of soda in a glass / cup. It would be very challenging for you, but it would be a fun project! Looking forward to what you are gonna make with this system. All the best, from the Netherlands.
This would be great for dispensing applications like color mixing, hydroponics, or selective oilers. Additionally, this system would benefit from a clutch mechanism or belt drive, which would greatly improve shifting time. Clutches would limit you to maybe 4 outputs, but a belt drive could be expanded almost indefinitely.
Reminds me of the Armatron gripper I had as a kid. It had only 1 motor to control a 6 axis arm via 2 joysticks. A modern version of it w/ steppers or even servo motors would be cool to see. I'd bet most of the design is available; just have to piece it together from various STLs.
Say tuning a transmitter amplifier together with harmonic filter and antenna matching network. Feed back to be supplied by frequency and phase discriminators as required. I think it would work well as there is usually a bit of back and forth tuning to get it spot on.
You could make an arm where the joints are controlled with strings and have this thing wind the strings :) also maybe you could have a bunch of branching tracks for the marble course and have this thing move back and forth and change the tracks like levers on a train track
I have made something similar in the past, it wasn't exactly a multiplexer like you made here, but it was designed to take one bidirectionnal input and have two unidirectionnal outputs. The way I made it was by using a sort of variant of a ratchet mechanism. The use I had for this mechanism was because I had a limitted amount of motors I could use (it was for a Lego mindstorms competition called Zone01, I was like 14 back then) But this allowed me to make two movements with one motor. And since we were limitted to 4 motors, this was a big plus
My sense is that the linear slide is the limiting factor. Some one already mentioned the use of a clutch, which is clever. I was thinking that you could have a rotary positioning system where you turn the drive to sequence it into position by stopping at the degree 0-360 that corresponds to that particular output. Or better yet you could have the circular arrangement, but the drive motor is fixed, with each drive on its own linear slide oriented perpendicular to drive motor. Then you could activate each output by engaging it to the drive. The advantage here is not only do you have shorter engagement time for the ouput, but you also have the possibility of engaging multiple outputs at once if they are needing to be turned in the same direction.
Not sure what you'd use it for, but I had an idea for an improvement to the mechanism. Right now you have the 3 blue gear + worm parts that mesh with the moving motor's gear. You could move those up to make room for another gear between them and the motor. The 3 additional gears would be longer, with only a small gap between each one, so that the motor has the option of stopping halfway between 2 positions and engaging both at once if 2 of them need to go the same direction.
Hello, I think your concept is good. If the individual gears could be coupled to the main drive at any time via a servo, then several axes could also be moved at the same time. You only have to pay attention to the maximum torque of the axes, otherwise the motor will stop.
This would also be really useful in hvac ducting, you could measure each rooms temperature and redirect heat/cooling to whatever room in your house that needs it the most.
I feel like the carriage system would be redundant unless it could have another purpose of its own - how about an automatic media flow discharge system - each wormed section has a container that takes an inflow of something like small ball bearings or water, and the carriage could be programmed to move to the container with the greatest mass to empty and transport its contents to an outflow or collection of some sort. Could be used for harvesting growing materials depending on their state of growth, too.
I've had some display ideas over the years that wanted something like this. I figured it would be long rods with magnetic clutches, so the things that needed driving could grab on until they were where they wanted to be. I had an idea, but not enough motivation/time/money, to make a grid wall of boxes, square from the front, and deep, with a roller at the back, an open front, and a smooth platen behind the front, held on by one side only. A loop of some kind of paper or plastic would loop around the back roller, and over the front platen, and it would be divided all along its length into colored squares, say 16 of them - good computery number. Spinning the roller would slide the tape over the platen, revealing different colors. A wall of those would make a low-color display, and changing to the next image would be a long, artsy process of 1 big motor spinning a vertical rod, and various magnetic clutches engaging particular row rods via gears, with more magnetic clutches engaging particular rollers to gears on those row rods. Seems like something your endless patience could pull off :)
My idea for how to use this is sort of a tentacle robot. Each joint along the tentacle can be controlled with a wire (like a tendon) that's tensioned or released, and we can do this from one motor using this mechanism. It could then also do some prioritising to get the segments near the base of the tentacle in their right position first, since they affect the position of all the higher up tentacles. Or do it the other way around so it doesn't smack something with the full length of the tentacle while trying to get its position right. The advantage of using a mechanism like this instead of individual motors winding up wires is that you can have a huge number of segments. Instead of doing this with a rigid robot and wires, you could of course go back to the soft robot idea with inflatable segments and the motor just switches between which segments to inflate/deflate. Could be useful for like, a worm that moves along the ground by inflating one part and deflating the other, but that's probably going to be a pain to make work just right.
This brings me back to my days of printer repair. More complex laser printers will have a bare minimum number of motors, with systems to engage and disengage different drive sections to accomplish different tasks, like feeding paper from the manual input or duplex printing. A common mechanism is a gear with a few teeth missing. It will usually have a slope that is engaged with a spring, and a key being held back by a solenoid. When the solenoid is fired, the gear will be free to advance (with the spring causing it to rotate enough to engage the teeth) and the gear will do a complete rotation before coming to rest against the solenoid. Another mechanism is having two different gear trains coming off the same motor, with a T rocker to engage only one train in one direction, and the other train in reverse.
The classic application for a system like this is a vending machine. Especially the kind where stacks of identical items are stored between a pair of screw augers, when you choose an item one single motor engages with the chosen auger and turns it, I think in most vending machines they use common shafts and some electromagnetic clutches to decide which shaft gets engaged, but I have observed at least one fizzy drinks machine that uses a 2 axis picker that moves in front of your chosen drink, uses a little plastic Allen key to spin the auger, the drink gently moves on to its platform and then it can move itself to the window for a handoff to the customer.
I was thinking of something like this for a project i had in mind. I was thinking of something like a car transmission system where one input can control multiple outputs. The project i had in mind wad multiple pop up shelves which will stay hidden in their default state and will pull out vertically when needed. It's too complicated for me to attempt as an actual shelving project for my home but the concept holds true. This mechanism will allow the user to have multiple pop up shelves at a bare minimum cost
You could add a latching mechanism. A type of hinged/spring loaded clip that engaged in the output shaft gears to keep it in place while the drive motor is not in place. It would prevent any drifting/back driving, and maybe help with gear meshing. The drive motor could lift it out of the way when engaged. Just an idea for an added feature. Not really sure if it's necessary for all applications.
Nice concept. it got my creative juices flowing, as i am typing i am getting more and more ideas. i think that it could benefit from closer spacing between output gears from faster switching. it might also be fun to look at a pneumatic switching system this could make the switching between output gears much faster (possibly with latches for positioning). i think this system has a few drawbacks 1. the outputs are free to be back-driven when the drive gear is not meshed. 2. during the initial stages of the meshing the gear load is very high. 3. is it slow. but these are all issues which can be addressed.
Reduced weight for the same functionality has space equipment written all over it. A rover that only does one thing at a time anyway has plenty of time to switch between drive outputs. Also, If you put geared outputs around the drive axis, you can condense the outputs. Then bring your output shafts out to the end of the mechanism, and stagger the lengths so you can put gears or pulleys on the outputs and keep it in line with the rotation axis. This would help reduce bulk, and be suited to being fixed along a segment of, say, a forearm for controlling wrist movements. Wrists don’t have a large range of motion, but encompass many axis into one joint. Many string-and-spring driven joints would work well in such a small area. You should add a sensor on the drive motor for feedback. Maybe 2 sensors: one geared for a finer resolution. That multiplexer you made for Ultron got me super curious too, I can’t wait to see what you can come up with for this mechanism!
I think it becomes more useful the closer the input gears are spaced but the further the outputs are spaced. If you had a large matrix of silos and needed to fill one of them at a time from an incoming truck, a system like this could allow one big engine to control 24 conveyor belts across a very large area. if you specifically need 3 outputs, you should put them in a ring configuration so at most you only have to move one worm gear distance.
Did you know that if you push a positive taper into a matching negative taper and twist, they'll lock together? Maybe not as reliable with plastic, but the concept is used even in high torque applications like vehicle transmissions and mill collets. I'd drop the sliding motor and instead opt for a linear+rotary shaft, commonly found in drill presses. Have your positive taper trapped between pneumatically actuated "shift forks," to actuate the positive taper into and out of a gear with a matching tapered bore, allowing you to quickly engage and disengage that output, or have multiple outputs engaged at once. You could add a second shaft for reverse.
Use it to drive a series of cable drums, each corresponding a sheet on a sailboat, and build a self sailing sailboat drone. One motor means lower weight in the boat = better performance. Create linkage to move motor down low for ballast. Using only one motor means less power draw, means enhanced duration.
Suggestion for you - Remove the lead screw setup and replace it with a fixed rotating shaft with straight cut gears at each power take off position. Have a hinge at the end of each worm gear bearing so that the worm/straight gear assembly and be pulled/pushed down by an electromagnet/pneumatic source to engage the output. The worm gears will still mesh fine at a different angle. This will mean multiple simultaneous drives can be used, reduce parts cost, increase switching speed, remove lag between engaging and disengaging gear sets and enable outputs to be selected without moving other outputs on the way as is the case in the current setup. Really cool idea though, hadn't thought about that.
One idea I had for the gearing efficiency is output overlap, where the drive gear goes to worm gears that have multiple internal outputs which in turn drive the system outputs. Thus having 3 outputs come from 8 internals which come from 1 drive input.
A consideration is that initial meshing of the gears, if the driving motor was moving and there is a load on the gear it’s trying to drive, all the weight would be on the tip of the tooth that meshes. So for a system you have here it works because it waits until full engaged to move.
Could be used as drive for solar tracker gimbal. Use several in concert as stacker robot for kinetic battery tower. Use it to control I/O valves of a fluid manifold that needs conditional balancing or differential.
Interesting, I've been thinking about this for a while. Unfortunately I don't have a workshop. The most interesting concept, that maybe you could try, is to have a bistable latch that engages the outputs, with a magnet. If you pass another magnet at high RPM, the magnets won't have time to connect, therefore the latch doesn't switch, if you "slow down" the magnets will switch the latch and engage the output. I didn't have the time to prototype the design yet, but I imagine that with a compliant material design on the latch/output gear, it may be able to produce a very fast multiplexer.
*Regarding the priority selection:* Both of the problems you had with your current prioritization scheme can be solved if you weight the errors by the distance the motor is from the output. (Score[n] = Error[n]/Dist[n]) You just need to check for one special case, (0 divided by 0 should be set to equal 0 not #NaN), #INFs from division by zero are fine. The scale you use for distance is tune-able to make the motor either prioritize bigger errors, or closer errors. In the ideal case it would be set based on the time it takes to travel vs the time to rotate the pot. *Regarding outputs:* Outputs that only need one-directional drives, can be paired together with a differential and two ratchets to be alternately drive-able based on the direction the motor turns; no movement necessary. (Super fast multiplexing.) I've used these techniques myself with lego robotics. (I was poor and couldn't afford to buy extra motors.)
It would be very interesting to see the linear selector swapped with a rotational output selection method. You may lose the ability to scale upwards to more than three outputs, but it would allow you to pivot between driving any of the three outputs to any other output instantly. I wonder if the system might allow for two outputs to be enabled at the same time or if that "feature" would force you to switch gears more slowly to avoid dual-driving the outputs. The rotational selector should ideally allow you to intentionally dual-drive outputs as a way of starting the next prioritized action while finishing the current action. Food for thought!
Cool! It seems like it'd be more efficient and smaller if the motor itself didn't move, but maybe a gear sliding along the shaft or even some kind of clutch mechanism.
That's how older milling machine used to work (before the advent of cheaper/smaller electric motor). They used a single beefy, normally 3 phase, motor and multiplex it to the 3 axis and spindle via clutches.
I think this could make a great addition to the open dog. I can see this being a sort of backpack for the dog making a cable driven arm, add a forth position and you have a grip control. could be a neat application
You could use the elevator algorithm to ensure starvation doesn't happen. The requirements of this are very similar to hard disk read scheduling when it had to scan the read head across the media. The elevator algorithm is to keep moving in the same direction while there are more things to do in that direction and stop at anything you encounter along the way that needs to be fixed. So it never needs to backtrack and it is guerenteed to eventually get to every motor and have a good latency for almost all load patterns. It's a good sweet spot for this type of optimization problem. It's named after how an elevator operates, it doesn't go all the way to the destination of the first pressed button, it stops at all floors it needs to along the way to that destination even if the passengers entered later.
I teach First Lego League (FLL) to 5th through 8th graders. The robot building competition allows only using one controller which can only control 4 motors and 4 input sensors. I showed them this same concept three years ago on using two motors to activate any number of discrete output shafts. I did it in a very similar way using worm gear drives so those discrete output hold their position when not being driven. I used multiple drive gears coupled to the main drive motor so the movement of the drive system was minimized. Center position drives the center gear. Sliding one gear width left of center drives the output shaft left of center. two gear widths left of center drives the second output shaft left of center. One gear width right of center drives the output shaft to the right of center. The separation of output shafts is determined by how many output shafts are incorporated into the system. This also allows output shafts that are side by side to be run simultaneously by moving the drive gear in half gear distances so the drive gears engage half of each of two driven gear trains. The worm gear drives were more complicated than yours in order to have the driven gears all side by side in a row. I accomplished this by running the worm gears with a bevel crown gear that meshes with the side of the driven gears.
Maybe you could use it in the Great Ball Machine to drive a grab lift? The operations involved in moving the ball (send carriage to ball input, lower grab, close grab, lift grab, send carriage to ball output, lower grab, open grab) can take place one at a time, so fits the mulitplexer perfectly with three axes (carriage position, grab height, grab open/closed)
This mechanism would be useful for a number of niche systems that I think you could make a great robot to accomplish those tasks. A seed planting robot that works in a prepared garden bed to separate seeds from a bulk supply, load them into a mechanism that can be activated in sequence by mechanical multiplexing as needed. A similar idea could make an art robot that uses colored sand and planned dispensation to make beautiful robotic art based on an image. I love your work, and you make my brain wiggle when I think about what you do.
Using a weighted function for prioritization of the outputs may be better than having these many conditions, something like: priority = (difference^2) / distance Will stay at an output until it is fully corrected (because distance is close to zero, therefore the priority approaches positive infinity), and will move to correct the next output, but taking the distance in consideration. This means that a small difference for a closer output will be corrected before moving a large distance.
Interesting question! One application might be a robot or vehicle which needs to "transform" or unfold such as in opening hatches, doors or changing modes (like an automatic convertible car roof). This doesn't happen frequently and doesn't need to have each drive output run at the same time. For robots, maybe the arm drive and the motion drive are aways done separately? I guess the key is applications most suited do not require simultaneous driving, but rather very low frequency switching demands between output gears. Also there must be a big cost difference between the gear box selector motor and the drive motor or if they are equal, there needs to be at least 3 drive outputs or you'd be better off just driving two outputs with the two motors. Also this application benefits when motors are especially expensive compared to the cost of building and maintaining the mechanism, or if the distribution of motor weight is of primary concern. I've noticed alternative mechanisms in modern low-priced printers where motors are used for multiple purposes, the difference being the mechanical couplings effectively concatenate such that beyond a certain physical limit, a coupler disengages the first mechanism which allows the motor's continued running to go on to effect another mechanism. So the same motor feeds the paper until the page is through and thereafter opens a hatch and draws the next page or whatever. Rather than a general purpose automatic gearbox, it is an application-specific linear concatenation of coupling effects. It's hard to say more than this in the general case when it might be said that you have a solution looking for a problem.
You could also have the motor run two knobs at the same time. In this build, with a longer gear, it could turn two knobs in the same direction until one of them reaches the correct position, then slide over and finish turning the other one.
For something that moves that slowly, it can be applied where time isn't critical, like a mechanical scoreboard changer at a baseball game. A long shaft that can change all the innings and total scores for each team, controlled remotely by the umpire.
Possible improvement for the programming part: instead of only considering the largest angle to decide which motor is prioritised you could also include a penalty for travel time. With travel time I mean the time it takes for the driving gear to move to the next gear. For example to get the highest score you could write something like "score = angle_deviation - K*travel_time" or "score = angle_deviation/(1 + travel_time)". Doing it this way has two advantages. Firstly you don't need additional code to ensure the motor finishes whichever gear it is currently doing. The current gear has travel_time==0 so it will always finish the current gear. Secondly in the current setup it is possible to have the scenario with priority left>right>center. The motor will first go to left, then to right and finally to center. It is much quicker to do left -> center -> right. Only when the center gear has a very small deviation would it be sensible to skip over it.
Small items or specific type of item sorter. Add a sort of scanner system or mechanical separator that allows you to use this to select which screw you want.
you could switch a robot between operating modes. in a rover-style robot you could for instance use the same set of motors to alternately drive and operate some sort of manipulator/arm. or as an iteration of this: create a parallel mechanic multiplexer to drive all leg joints of hexapod-style robot simultaneously, and iterate over each leg in turn
I could see something like this being integrated in some factory machines. Maybe you could make a robot foosball table. I love living in the US because the announcements always say Tuesday and it shows up monday night. Im spoiled for that. Awesome video!!!
@@tippyc2 With a fast motor and a can do attitude it might be fun. Maybe not the best possible design, but It would still be a cool use of the mechanism.
You could use it as a three axis alignment corrector, so if a plane is off kilter and at the wrong height it will correct these errors fixing the most egregious first. Slow stabailasiation for sensors that would produce noise if corrected quickly but could be corrected for slowly - accelerometers, mems, maybe sonar
I really love the idea of a multiplexer in space rovers! You can move multiple joints with just a single motor, it would save a ton of weight! Or when it comes to reliability, I wonder if there is a way to create a multiplexer with multiple motors. Like... join A and join B are controlled by motor A and B respectively. But if motor B was to fail, motor A could use a multiplexer to drive both joint A and joint B. It would make the movement process slower, but you would still be able to move a joint that you wouldnt be able to otherwise. I would really love to see you develop something like that!!
If you know the position of the input gear, then you can avoid the need for a synchro. Just need an encoder on the input gear and then store the position when it disengages a gear and go back to that position when it comes back to the same gear.
Watching 3d printers gets me excited for the future where 3d printers will be much more accessible and much quicker. I know paper printers aren't an exact analog for 3d printers but it gives my mind something to visualize when comparing how much better technology can grow
You could use it to do an IOT weather and other data gauges. You could have all sorts of pointers like tempature, humidity, wind speed, wind direction, rain fall, fire danger, un read emails, steps taken, heart rate, stocks, soil moisture, days since last accident, uv light, 3d printer progress, what ever else you could think of. All those things don't have to be updated continuously instantly they can be updated when convinent and the multiplexer goes around and adjusts the values as they come in some times once a day or once a hr, or every minute for some
If there's sufficient load on the outputs, the slim sections of the yellow and blue spur gears are going to be under a lot of force as they start to mesh. Might be worth only running the main drive motor once it's in position to allow it to mesh completely
I recommend using sun and planet gears. you could have all motors sun gears run continuously by a motor and all you have to do to actuate the output (the arm) is grip the ring gear with some break pads. You could even actuate multiple outputs at once (in the same direction).
You could use it to build a crane. You would only need one motor for all the axis, and you could prioritize the base rotation first for example, so the crane is never in the wrong spot while lowering the end.
This would be a great idea to control a cable-driven appendage such as a tentacle or tail. The footprint of the Multiplexer would allow it to be mounted on a rail system that could be driven by a worm gear the length of a robot. This could effectively give opendog an extra appendage to interact with its environment. It could even play fetch at that point.
I am interested in the opposite idea as well: multiple motor or gear ratios actuating one thing. Then if you could have big fast movements or fine slow precision on the same leg
Output shaft to cable shaft or tensioned pulling system.. Change from screw driven selector to planetary gear selection with orbiting gears engaging such that more than one can be engaged based on the available power measured perhaps as back current. Use the arm to create a 3D printer or clay sculptor or brick layer.
It should value distance, not just the biggest difference. So it'll go toward a nearer output rather than the larger one. So if it's working on the left one and you change the other two, it'll start moving to the middle. But if you immediately change the left one some more, as long as it's closer to the left than the middle, it'll turn back around. If both are near equal, for example it's working on the middle and you change the sides, it should choose based on the difference from input/output as a tiebreaker. The way you set it up it seems unintuitive for it to move past the middle output to the other side.
It looks like there's enough space to fit three more gear-screws rotated 180degrees to run three reverse direction outputs in the offset space between the current gear-screws.
It might be a good idea to use a clutch system with permanently meshing gears. That would allow you to switch outputs quicker and address multiple outputs at once if they all need to be turned in the same direction
The clutches could be controlled by cheap servos, so that should work work well.
@@clonkex Or better to use magnetic clutch systems to get better response times
I prototyped such system for a loading/unloading filament mechanism, using small clutchs from copiers. The issue is they went hot when energized, and melted my PLA parts :o/ I may try again with ABS, one day.
This has been used in colour laser printers for many years; magnetic clutch to drive each colour drum in turn from one motor
@@Andrew_Sparrow That's really interesting. I'll have to look for more information on it.
I feel like this could be used for a self leveling platform. A table would be the easiest thing to say, you can plop it anywhere and the table will adjust the leg height automatically to become level.
or a theodolite or dumpy level laser
I've been trying to think of a design for a washing machine stand that I can level more easily than adjusting the feet of the machine. This mechanism is better than any of the designs I had in mind.
This would be very useful for auto-bed levelling for 3d printers.
You could use this to keep soil irrigated for a 3-dimensional stack of plants. Moisture sensors could tell you which plant needs water then this could be used to send the dispenser to the proper xyz position.
Plant vending machine! I like it :)
Wonderful idea! But the idea of a moisture sensor would be troublesome from the perspective of establishing values. Thus could be bypassed via a little chip that gets programmed with set intervals, or maybe a knob that adjusts the timing by 5% every revolution.
@@garnerday7149 farmers just use drones or quadcopters with cams for multi-spectral images. You don't really need a boots-on-the-ground solution for that other than in remarkable spots. Some trained sw and that kind of data would handle much of what could be automated.
Or just run a tube to each level and select the pump output with a solenoid. This thing is massive overkill for such a simple application
This would be great for sorting or dispensing discrete amounts of items.
Yes. For dispensing ingredients or pigments.
Next video: "James' Marvellous Cocktail Machine." lol
I really like the idea of combining these with peristaltic pumps for hydroponics... or a home bartender robot.
maybe it could be integrated into the really useful robot build as a base station or something
Perhaps it could be incorporated into a robot that makes funnels ?
I think it would be practical to use it for controlling a lot of water valves.
yes, or air valves, like for pneumatic soft robots!
Yes, and also multiple perestaltic pumps, the most awesome water valve of them all XD
Anyway... retrofitting existing water valves would save them from the trash! They can be quite demanding in torque!
Multi tap beer dispenser for a robotics event after COVID?
Robot bartender? Watering a bunch of plants? Might be interesting if you could make the multiplexer circular. That way you’re never more than half the outputs away from your next position.
@@Wileama this one is great because unlike a soft robot, no more than one output would need to be actuated at a time
Does James even sleep? How does he engineer and build so much stuff?
I know right? I take so long to design things in fusion360 and print them. Then when I do, it's usually wrong, so I have to try again!
practice and experience?
@@BlueScreenCorp and sponsors that give printing supplies ! x']
even James can get a lot of money running this channel alone, I think he would rather do it with a team and share the revenue
Он спит, пока течёт его любимый кетчуп
Или печатается очередная деталь
Maybe it would be useful for a robotic bartender, it operates many pumps that funnel into a glass?
AvE has entered the chat
Tamás Fehér would James have to 3d print the funnel ?
@@travisash8180 As far as I know FDM 3d prints are rarely watertight out of the printer, so it might be better to use a kitchen funnel from a dollar store or something like that
Maybe as a novelty in a home bar. The "one at a time" control limitation is gonna make that super slow.
It reminds me of those useless boxes that turn off their own ON switch. You could make the ultimate "Useless Box"
yeah, like a useless box with three analog input/output instead of only one digital one.
Yesss this was my thought also!
This can be used in a flip clock.
It will rest on the 'seconds' position, but will move along when the 'hour' or 'minute' change.
Could even be extended to a calendar or any other kind of counter.
It's titles like that which bring a smile to my face because i know I'm going into the video knowing absolutely nothing
But come out knowing quite a bit.
Sailboats often have an assisted motor for pulling anchor chains, hoisting sails and adjusting the sail position, people usually manually change the input to what they need.
It’s awesome that you share these videos, but the even bigger thing is that you provide a literal master class in critical problem-solving, as it relates to identifying all case use scenarios. This video is a great example - you walk us through, step by step, how you address each potential scenario and subsequently beef up your code to resolve the issue in question. This sort of critical logic application is less common than we might collectively hope for. 😂
As an aside, my daughter is 8 and I promised her we would build one of your robotic dogs - looking forward to the challenge!
Shame about the lack of printing music these days. Always enjoyed those bits in the past. The cute chiptune'y music made me smile
People like you are what keep me on TH-cam.
I recently did a project very similar to this, but with few differences. The main motor was connected to a long lantern gear, mounted to motor shaft. Corresponding star gears were engaged to this lantern as the outputs. These star gears were mounted on a component, which featured solenoid, with mounted gear. When the solenoid was activated, whole mechanism engaged and torque was transfered. Advantage of this approach is that we can have more than one output engaged at the time.
I didn't see it at first, you did a great job presenting every step....then it clicked. Nicely done. Could you add two more axis, a joystick(or integrate like a game controller) and an arm with a cup? Excavator, or those could be the programing inputs for like a pick and place arm, safe cracker, laser guided drill press, etch a sketch with retractable pen....damn the possibilities.
Have your TV or PC screen oriented at optimal position. With the proper sensors it could follow you around the room. This would be awesome in a home gym where one has multiple exercising machines spread throughout the room or while working on a project in the garage/shed.
this would be great for some kind of sequence of events, like a "loading" robot that first grabs a ball/projectile, then swings a arm to a receiver then releases the clamp and then shuts the receiver hatch. after it will then reset all items to their original position
A trebuchet robot! Arm, load, aim, fire!
Electromechanics. Technology Connections has a couple great vids on a jukebox with just this type of idea
@@brendandor I'm thinking phase 1 a catapult winding back, phase 2 something dispensing or an arm dropping some type of projectile in, and phase 3 release to fire at a predetermined position like the entry way to your room just for fun
If you used a double clutch system on each output, you could have a motor thats turning at a constant speed, and outputs triggered simultaneously in either direction 🙂
Made something like this in college. It had twelve outputs that were vertical screws driving an indicator. The planned use was to indicate the tide for the next twelve hours.
You could turn arround the first worm gear so that there is less distance between the three input gears. That reduces travel time, espacially between the first and the second input. Then you could hook up the most often used output to the worm gear in the middle, and the second most often used output to the first worm gear.
Super Vorschlag
This seems almost identical to a manual gearbox, just with multiple outputs instead of a single output shaft and a moving motor instead of clutches.
Rather than moving the motor and using the gears as a clutch, all the gears and motor would stay fixed and meshed, and clutches would engage or disengage each output. This would also have the advantage of using multiple outputs at once rather than being stuck with one at a time. Plus much faster engagement.
With friction clutches it would make a good tracked vehicle drivetrain, or 8x8 wheeled vehicle. Engaging one side or the other to steer, and if using wheels only engage the wheels that are on the ground.
My suggestion, build a second one and have its output control the pots on the first one, and vice versa.
This mechanism remembers me so much of the Prusa MMU, it's basically the same exact concept for multiplexing multiple filament to a single filament extruder.
The absolute best use of this device is for TH-cam video content, and showcasing the possibilities of 3d printing.
So I think you nailed it first try.
Or a coding challenge :-)
When I was a kid I had a Radio Shack Armatron. A whole robotic arm that ran off of one drive motor and lots of clever gearing.
I still have mine somewhere. Very clever system of gears, and 4 of them were dual-speed too!
I have my Armatron on my desk. Doesn't work too well anymore, but always reminds me of my love for robotics that I've had since I believed in Santa.
I really wanted one but was poor... making a replica would be a good series.
This is such a lovely thing... it's so hypnotic just watching it slide back and forth turning stuff
It seems like if you have to sweep the entire distance anyway, you can process the middle output partially on its journey across.
I love how you show the iterative modification to the logic.
Mechanical multiplexer is an amazing idea. It is a perfect solution for a robotic arm as each degree of freedom needs high torque and high precision = big and expensive motor. Unused channels would require breaks though.
I am looking forward to the future videos on this topic.
There is a lot of knowledge and know how behind making something working as smoothly as shown.
It might look easy but it is certainly not. I appreciate the quality of things on this channel.
I've seen this mechanism at a huge industrial greenhouse. A single motor on a rail would open or close dozens of windows in different patterns.
Makes me wonder if we can benefit from multiplication. 3 data points can control 7 different variables, 3 completely. 4 can control 14, 9 completely. (Math done in head while exhausted at 1:24am) There should be a way to mechanically pull it off.
Example, say gear A, B, C, and D. AB, AC, AD, BC, BD, CD, ABC, ABD, and BCD can be controlled pretty accurately while varying from each other by just the 4 inputs. It costs a bit of accuracy, but can increase the amount of control through minimal gears. And accuracy increases as precision does.
Hope this makes sense. Really worn out ATM.
James: Builds SLAM-located, AI navigating quadrapedic robot dog with ROS and advanced power electronics.
Also James: "So I just used tape and zipties to attach these potentiometers."
That's the expertise of a true engineer!
"anyone can design a wall hook, but an engineer can design a wall hook that is just as strong as neeeded"
@@jearlblah5169 And then the engineer slaps a 5x factor of safety on it anyway because people are stupid
@@tippyc2 but also, an overly tight tolerance that fails when imprecise installations come into play
This is used in high school robotics commonly for alternative modes of robot movement. So in the case the robot has to climb there's really no point to also be able to drive at the same time. So it could switch between drive and climb as once lined up for climbing the drive doesn't matter.
Hi James, as a LEGO fan I didn’t have that many motor, so I used this system of a sliding motor to many outputs a lot. I had a good inspiration with Technic set 8082 Multi Control Set. You could build 4 models with this set and one of them was a skid steer forklift vehicle with this system. I used it the control a self designed crane to move the boom, arm, hook and rotate the crane. With 2 motors and 2 up and down joysticks I could move a model with 4 functions. This was absolutely great. Now I have a hole draw full of motor so I don’t use this system any more, because it is very bulky and only works in large models. If you interested in the LEGO set, there are many website that have LEGO instructions to watch online or download in pdf. It is quick a interesting model, because it is very compact and the positioning motor only moves the rack 1 stud per function. So to compact this system you could learn a lot from this model :)
You could build a rover that moves forward and back, slide and it turns left and right, slide and the arm moves up and down, slide and it opens or closes the grabber. With rotation sensors you could program it to move a programmed path or with a light sensor make it a line following rover. Make it pick up objects or even give it a fifth function os it can alternate between different attachments. Or rotate the grabber so it can pore a can of soda in a glass / cup. It would be very challenging for you, but it would be a fun project!
Looking forward to what you are gonna make with this system.
All the best, from the Netherlands.
This would be great for dispensing applications like color mixing, hydroponics, or selective oilers. Additionally, this system would benefit from a clutch mechanism or belt drive, which would greatly improve shifting time. Clutches would limit you to maybe 4 outputs, but a belt drive could be expanded almost indefinitely.
Reminds me of the Armatron gripper I had as a kid. It had only 1 motor to control a 6 axis arm via 2 joysticks. A modern version of it w/ steppers or even servo motors would be cool to see. I'd bet most of the design is available; just have to piece it together from various STLs.
Say tuning a transmitter amplifier together with harmonic filter and antenna matching network. Feed back to be supplied by frequency and phase discriminators as required. I think it would work well as there is usually a bit of back and forth tuning to get it spot on.
You could make an arm where the joints are controlled with strings and have this thing wind the strings :) also maybe you could have a bunch of branching tracks for the marble course and have this thing move back and forth and change the tracks like levers on a train track
Like the string idea, there is so much potential with that. Like puppets (Pinocchio), or the wooden Village puppets from Shrek.
I have made something similar in the past, it wasn't exactly a multiplexer like you made here, but it was designed to take one bidirectionnal input and have two unidirectionnal outputs. The way I made it was by using a sort of variant of a ratchet mechanism. The use I had for this mechanism was because I had a limitted amount of motors I could use (it was for a Lego mindstorms competition called Zone01, I was like 14 back then) But this allowed me to make two movements with one motor. And since we were limitted to 4 motors, this was a big plus
My sense is that the linear slide is the limiting factor. Some one already mentioned the use of a clutch, which is clever. I was thinking that you could have a rotary positioning system where you turn the drive to sequence it into position by stopping at the degree 0-360 that corresponds to that particular output. Or better yet you could have the circular arrangement, but the drive motor is fixed, with each drive on its own linear slide oriented perpendicular to drive motor. Then you could activate each output by engaging it to the drive. The advantage here is not only do you have shorter engagement time for the ouput, but you also have the possibility of engaging multiple outputs at once if they are needing to be turned in the same direction.
Not sure what you'd use it for, but I had an idea for an improvement to the mechanism. Right now you have the 3 blue gear + worm parts that mesh with the moving motor's gear. You could move those up to make room for another gear between them and the motor. The 3 additional gears would be longer, with only a small gap between each one, so that the motor has the option of stopping halfway between 2 positions and engaging both at once if 2 of them need to go the same direction.
Hello, I think your concept is good. If the individual gears could be coupled to the main drive at any time via a servo, then several axes could also be moved at the same time. You only have to pay attention to the maximum torque of the axes, otherwise the motor will stop.
This would also be really useful in hvac ducting, you could measure each rooms temperature and redirect heat/cooling to whatever room in your house that needs it the most.
I feel like the carriage system would be redundant unless it could have another purpose of its own - how about an automatic media flow discharge system - each wormed section has a container that takes an inflow of something like small ball bearings or water, and the carriage could be programmed to move to the container with the greatest mass to empty and transport its contents to an outflow or collection of some sort. Could be used for harvesting growing materials depending on their state of growth, too.
I've had some display ideas over the years that wanted something like this. I figured it would be long rods with magnetic clutches, so the things that needed driving could grab on until they were where they wanted to be. I had an idea, but not enough motivation/time/money, to make a grid wall of boxes, square from the front, and deep, with a roller at the back, an open front, and a smooth platen behind the front, held on by one side only. A loop of some kind of paper or plastic would loop around the back roller, and over the front platen, and it would be divided all along its length into colored squares, say 16 of them - good computery number. Spinning the roller would slide the tape over the platen, revealing different colors. A wall of those would make a low-color display, and changing to the next image would be a long, artsy process of 1 big motor spinning a vertical rod, and various magnetic clutches engaging particular row rods via gears, with more magnetic clutches engaging particular rollers to gears on those row rods. Seems like something your endless patience could pull off :)
My idea for how to use this is sort of a tentacle robot. Each joint along the tentacle can be controlled with a wire (like a tendon) that's tensioned or released, and we can do this from one motor using this mechanism. It could then also do some prioritising to get the segments near the base of the tentacle in their right position first, since they affect the position of all the higher up tentacles. Or do it the other way around so it doesn't smack something with the full length of the tentacle while trying to get its position right. The advantage of using a mechanism like this instead of individual motors winding up wires is that you can have a huge number of segments.
Instead of doing this with a rigid robot and wires, you could of course go back to the soft robot idea with inflatable segments and the motor just switches between which segments to inflate/deflate. Could be useful for like, a worm that moves along the ground by inflating one part and deflating the other, but that's probably going to be a pain to make work just right.
This brings me back to my days of printer repair. More complex laser printers will have a bare minimum number of motors, with systems to engage and disengage different drive sections to accomplish different tasks, like feeding paper from the manual input or duplex printing. A common mechanism is a gear with a few teeth missing. It will usually have a slope that is engaged with a spring, and a key being held back by a solenoid. When the solenoid is fired, the gear will be free to advance (with the spring causing it to rotate enough to engage the teeth) and the gear will do a complete rotation before coming to rest against the solenoid.
Another mechanism is having two different gear trains coming off the same motor, with a T rocker to engage only one train in one direction, and the other train in reverse.
The classic application for a system like this is a vending machine. Especially the kind where stacks of identical items are stored between a pair of screw augers, when you choose an item one single motor engages with the chosen auger and turns it, I think in most vending machines they use common shafts and some electromagnetic clutches to decide which shaft gets engaged, but I have observed at least one fizzy drinks machine that uses a 2 axis picker that moves in front of your chosen drink, uses a little plastic Allen key to spin the auger, the drink gently moves on to its platform and then it can move itself to the window for a handoff to the customer.
I was thinking of something like this for a project i had in mind. I was thinking of something like a car transmission system where one input can control multiple outputs.
The project i had in mind wad multiple pop up shelves which will stay hidden in their default state and will pull out vertically when needed. It's too complicated for me to attempt as an actual shelving project for my home but the concept holds true. This mechanism will allow the user to have multiple pop up shelves at a bare minimum cost
You could add a latching mechanism. A type of hinged/spring loaded clip that engaged in the output shaft gears to keep it in place while the drive motor is not in place. It would prevent any drifting/back driving, and maybe help with gear meshing. The drive motor could lift it out of the way when engaged. Just an idea for an added feature. Not really sure if it's necessary for all applications.
Nice concept. it got my creative juices flowing, as i am typing i am getting more and more ideas. i think that it could benefit from closer spacing between output gears from faster switching. it might also be fun to look at a pneumatic switching system this could make the switching between output gears much faster (possibly with latches for positioning). i think this system has a few drawbacks 1. the outputs are free to be back-driven when the drive gear is not meshed. 2. during the initial stages of the meshing the gear load is very high. 3. is it slow.
but these are all issues which can be addressed.
I really like thangs and hope it gets more popular.
Reduced weight for the same functionality has space equipment written all over it. A rover that only does one thing at a time anyway has plenty of time to switch between drive outputs. Also, If you put geared outputs around the drive axis, you can condense the outputs. Then bring your output shafts out to the end of the mechanism, and stagger the lengths so you can put gears or pulleys on the outputs and keep it in line with the rotation axis. This would help reduce bulk, and be suited to being fixed along a segment of, say, a forearm for controlling wrist movements. Wrists don’t have a large range of motion, but encompass many axis into one joint. Many string-and-spring driven joints would work well in such a small area. You should add a sensor on the drive motor for feedback. Maybe 2 sensors: one geared for a finer resolution.
That multiplexer you made for Ultron got me super curious too, I can’t wait to see what you can come up with for this mechanism!
I think it becomes more useful the closer the input gears are spaced but the further the outputs are spaced. If you had a large matrix of silos and needed to fill one of them at a time from an incoming truck, a system like this could allow one big engine to control 24 conveyor belts across a very large area. if you specifically need 3 outputs, you should put them in a ring configuration so at most you only have to move one worm gear distance.
the dog food robot from "back to the future" 3 mechanisms (can drop, arm move, and open can) if you want arm twist too.
Did you know that if you push a positive taper into a matching negative taper and twist, they'll lock together? Maybe not as reliable with plastic, but the concept is used even in high torque applications like vehicle transmissions and mill collets.
I'd drop the sliding motor and instead opt for a linear+rotary shaft, commonly found in drill presses. Have your positive taper trapped between pneumatically actuated "shift forks," to actuate the positive taper into and out of a gear with a matching tapered bore, allowing you to quickly engage and disengage that output, or have multiple outputs engaged at once. You could add a second shaft for reverse.
Use it to drive a series of cable drums, each corresponding a sheet on a sailboat, and build a self sailing sailboat drone. One motor means lower weight in the boat = better performance. Create linkage to move motor down low for ballast. Using only one motor means less power draw, means enhanced duration.
Suggestion for you - Remove the lead screw setup and replace it with a fixed rotating shaft with straight cut gears at each power take off position. Have a hinge at the end of each worm gear bearing so that the worm/straight gear assembly and be pulled/pushed down by an electromagnet/pneumatic source to engage the output. The worm gears will still mesh fine at a different angle. This will mean multiple simultaneous drives can be used, reduce parts cost, increase switching speed, remove lag between engaging and disengaging gear sets and enable outputs to be selected without moving other outputs on the way as is the case in the current setup. Really cool idea though, hadn't thought about that.
One idea I had for the gearing efficiency is output overlap, where the drive gear goes to worm gears that have multiple internal outputs which in turn drive the system outputs. Thus having 3 outputs come from 8 internals which come from 1 drive input.
A consideration is that initial meshing of the gears, if the driving motor was moving and there is a load on the gear it’s trying to drive, all the weight would be on the tip of the tooth that meshes. So for a system you have here it works because it waits until full engaged to move.
Could be used as drive for solar tracker gimbal.
Use several in concert as stacker robot for kinetic battery tower.
Use it to control I/O valves of a fluid manifold that needs conditional balancing or differential.
Interesting, I've been thinking about this for a while. Unfortunately I don't have a workshop.
The most interesting concept, that maybe you could try, is to have a bistable latch that engages the outputs, with a magnet. If you pass another magnet at high RPM, the magnets won't have time to connect, therefore the latch doesn't switch, if you "slow down" the magnets will switch the latch and engage the output. I didn't have the time to prototype the design yet, but I imagine that with a compliant material design on the latch/output gear, it may be able to produce a very fast multiplexer.
*Regarding the priority selection:* Both of the problems you had with your current prioritization scheme can be solved if you weight the errors by the distance the motor is from the output. (Score[n] = Error[n]/Dist[n]) You just need to check for one special case, (0 divided by 0 should be set to equal 0 not #NaN), #INFs from division by zero are fine. The scale you use for distance is tune-able to make the motor either prioritize bigger errors, or closer errors. In the ideal case it would be set based on the time it takes to travel vs the time to rotate the pot.
*Regarding outputs:* Outputs that only need one-directional drives, can be paired together with a differential and two ratchets to be alternately drive-able based on the direction the motor turns; no movement necessary. (Super fast multiplexing.)
I've used these techniques myself with lego robotics. (I was poor and couldn't afford to buy extra motors.)
It would be very interesting to see the linear selector swapped with a rotational output selection method. You may lose the ability to scale upwards to more than three outputs, but it would allow you to pivot between driving any of the three outputs to any other output instantly.
I wonder if the system might allow for two outputs to be enabled at the same time or if that "feature" would force you to switch gears more slowly to avoid dual-driving the outputs. The rotational selector should ideally allow you to intentionally dual-drive outputs as a way of starting the next prioritized action while finishing the current action.
Food for thought!
Cool! It seems like it'd be more efficient and smaller if the motor itself didn't move, but maybe a gear sliding along the shaft or even some kind of clutch mechanism.
You might want to look into electromagnetic clutches for this purpose, that would be interesting, too!
That's how older milling machine used to work (before the advent of cheaper/smaller electric motor). They used a single beefy, normally 3 phase, motor and multiplex it to the 3 axis and spindle via clutches.
I think this could make a great addition to the open dog. I can see this being a sort of backpack for the dog making a cable driven arm, add a forth position and you have a grip control. could be a neat application
You could use the elevator algorithm to ensure starvation doesn't happen. The requirements of this are very similar to hard disk read scheduling when it had to scan the read head across the media.
The elevator algorithm is to keep moving in the same direction while there are more things to do in that direction and stop at anything you encounter along the way that needs to be fixed. So it never needs to backtrack and it is guerenteed to eventually get to every motor and have a good latency for almost all load patterns. It's a good sweet spot for this type of optimization problem.
It's named after how an elevator operates, it doesn't go all the way to the destination of the first pressed button, it stops at all floors it needs to along the way to that destination even if the passengers entered later.
I teach First Lego League (FLL) to 5th through 8th graders. The robot building competition allows only using one controller which can only control 4 motors and 4 input sensors. I showed them this same concept three years ago on using two motors to activate any number of discrete output shafts. I did it in a very similar way using worm gear drives so those discrete output hold their position when not being driven. I used multiple drive gears coupled to the main drive motor so the movement of the drive system was minimized. Center position drives the center gear. Sliding one gear width left of center drives the output shaft left of center. two gear widths left of center drives the second output shaft left of center. One gear width right of center drives the output shaft to the right of center. The separation of output shafts is determined by how many output shafts are incorporated into the system. This also allows output shafts that are side by side to be run simultaneously by moving the drive gear in half gear distances so the drive gears engage half of each of two driven gear trains. The worm gear drives were more complicated than yours in order to have the driven gears all side by side in a row. I accomplished this by running the worm gears with a bevel crown gear that meshes with the side of the driven gears.
Maybe you could use it in the Great Ball Machine to drive a grab lift? The operations involved in moving the ball (send carriage to ball input, lower grab, close grab, lift grab, send carriage to ball output, lower grab, open grab) can take place one at a time, so fits the mulitplexer perfectly with three axes (carriage position, grab height, grab open/closed)
This mechanism would be useful for a number of niche systems that I think you could make a great robot to accomplish those tasks. A seed planting robot that works in a prepared garden bed to separate seeds from a bulk supply, load them into a mechanism that can be activated in sequence by mechanical multiplexing as needed. A similar idea could make an art robot that uses colored sand and planned dispensation to make beautiful robotic art based on an image. I love your work, and you make my brain wiggle when I think about what you do.
Using a weighted function for prioritization of the outputs may be better than having these many conditions, something like:
priority = (difference^2) / distance
Will stay at an output until it is fully corrected (because distance is close to zero, therefore the priority approaches positive infinity), and will move to correct the next output, but taking the distance in consideration. This means that a small difference for a closer output will be corrected before moving a large distance.
I feel like this could be used for a type of vending system. Or some kind of recipe builder.
Interesting question!
One application might be a robot or vehicle which needs to "transform" or unfold such as in opening hatches, doors or changing modes (like an automatic convertible car roof). This doesn't happen frequently and doesn't need to have each drive output run at the same time. For robots, maybe the arm drive and the motion drive are aways done separately?
I guess the key is applications most suited do not require simultaneous driving, but rather very low frequency switching demands between output gears. Also there must be a big cost difference between the gear box selector motor and the drive motor or if they are equal, there needs to be at least 3 drive outputs or you'd be better off just driving two outputs with the two motors. Also this application benefits when motors are especially expensive compared to the cost of building and maintaining the mechanism, or if the distribution of motor weight is of primary concern.
I've noticed alternative mechanisms in modern low-priced printers where motors are used for multiple purposes, the difference being the mechanical couplings effectively concatenate such that beyond a certain physical limit, a coupler disengages the first mechanism which allows the motor's continued running to go on to effect another mechanism. So the same motor feeds the paper until the page is through and thereafter opens a hatch and draws the next page or whatever. Rather than a general purpose automatic gearbox, it is an application-specific linear concatenation of coupling effects. It's hard to say more than this in the general case when it might be said that you have a solution looking for a problem.
You could also have the motor run two knobs at the same time. In this build, with a longer gear, it could turn two knobs in the same direction until one of them reaches the correct position, then slide over and finish turning the other one.
For something that moves that slowly, it can be applied where time isn't critical, like a mechanical scoreboard changer at a baseball game. A long shaft that can change all the innings and total scores for each team, controlled remotely by the umpire.
Prioritizing tasks is quite well understood in operating systems. You could investigate scheduling algorithms and use them in your multiplexer.
Possible improvement for the programming part: instead of only considering the largest angle to decide which motor is prioritised you could also include a penalty for travel time. With travel time I mean the time it takes for the driving gear to move to the next gear. For example to get the highest score you could write something like "score = angle_deviation - K*travel_time" or "score = angle_deviation/(1 + travel_time)". Doing it this way has two advantages. Firstly you don't need additional code to ensure the motor finishes whichever gear it is currently doing. The current gear has travel_time==0 so it will always finish the current gear. Secondly in the current setup it is possible to have the scenario with priority left>right>center. The motor will first go to left, then to right and finally to center. It is much quicker to do left -> center -> right. Only when the center gear has a very small deviation would it be sensible to skip over it.
Small items or specific type of item sorter. Add a sort of scanner system or mechanical separator that allows you to use this to select which screw you want.
you could switch a robot between operating modes. in a rover-style robot you could for instance use the same set of motors to alternately drive and operate some sort of manipulator/arm.
or as an iteration of this: create a parallel mechanic multiplexer to drive all leg joints of hexapod-style robot simultaneously, and iterate over each leg in turn
This would be super helpful in a hydraulic system such as an automated backhoe
I could see something like this being integrated in some factory machines. Maybe you could make a robot foosball table. I love living in the US because the announcements always say Tuesday and it shows up monday night. Im spoiled for that. Awesome video!!!
If he sped it up it could be used to make an amazing robotic foosball table. I really like this idea.
Wouldnt be viable, factory machines usually need to perform multiple functions at once.
It would work just not efficiently how bout that haha
@@tippyc2 With a fast motor and a can do attitude it might be fun. Maybe not the best possible design, but It would still be a cool use of the mechanism.
You could use it as a three axis alignment corrector, so if a plane is off kilter and at the wrong height it will correct these errors fixing the most egregious first. Slow stabailasiation for sensors that would produce noise if corrected quickly but could be corrected for slowly - accelerometers, mems, maybe sonar
I really love the idea of a multiplexer in space rovers!
You can move multiple joints with just a single motor, it would save a ton of weight!
Or when it comes to reliability, I wonder if there is a way to create a multiplexer with multiple motors.
Like... join A and join B are controlled by motor A and B respectively.
But if motor B was to fail, motor A could use a multiplexer to drive both joint A and joint B. It would make the movement process slower, but you would still be able to move a joint that you wouldnt be able to otherwise.
I would really love to see you develop something like that!!
If you know the position of the input gear, then you can avoid the need for a synchro. Just need an encoder on the input gear and then store the position when it disengages a gear and go back to that position when it comes back to the same gear.
Watching 3d printers gets me excited for the future where 3d printers will be much more accessible and much quicker. I know paper printers aren't an exact analog for 3d printers but it gives my mind something to visualize when comparing how much better technology can grow
This reminds me of the Armatron toy from many years ago in the sense that it uses a single motor to control many different motions. Very cool.
You could use it to do an IOT weather and other data gauges. You could have all sorts of pointers like tempature, humidity, wind speed, wind direction, rain fall, fire danger, un read emails, steps taken, heart rate, stocks, soil moisture, days since last accident, uv light, 3d printer progress, what ever else you could think of. All those things don't have to be updated continuously instantly they can be updated when convinent and the multiplexer goes around and adjusts the values as they come in some times once a day or once a hr, or every minute for some
If there's sufficient load on the outputs, the slim sections of the yellow and blue spur gears are going to be under a lot of force as they start to mesh. Might be worth only running the main drive motor once it's in position to allow it to mesh completely
I recommend using sun and planet gears. you could have all motors sun gears run continuously by a motor and all you have to do to actuate the output (the arm) is grip the ring gear with some break pads. You could even actuate multiple outputs at once (in the same direction).
This totally could reload balls that accumulate from various outputs with that great ball contraption you are working on.
You could use it to build a crane. You would only need one motor for all the axis, and you could prioritize the base rotation first for example, so the crane is never in the wrong spot while lowering the end.
I will use it to switch between different gear rastios for a robotarm. Thank you so much for this idea.
This would be a great idea to control a cable-driven appendage such as a tentacle or tail. The footprint of the Multiplexer would allow it to be mounted on a rail system that could be driven by a worm gear the length of a robot. This could effectively give opendog an extra appendage to interact with its environment. It could even play fetch at that point.
I am interested in the opposite idea as well: multiple motor or gear ratios actuating one thing. Then if you could have big fast movements or fine slow precision on the same leg
Output shaft to cable shaft or tensioned pulling system.. Change from screw driven selector to planetary gear selection with orbiting gears engaging such that more than one can be engaged based on the available power measured perhaps as back current.
Use the arm to create a 3D printer or clay sculptor or brick layer.
It should value distance, not just the biggest difference. So it'll go toward a nearer output rather than the larger one. So if it's working on the left one and you change the other two, it'll start moving to the middle. But if you immediately change the left one some more, as long as it's closer to the left than the middle, it'll turn back around. If both are near equal, for example it's working on the middle and you change the sides, it should choose based on the difference from input/output as a tiebreaker. The way you set it up it seems unintuitive for it to move past the middle output to the other side.
I think this would be great for a peristaltic pump setup for watering or treating plants
Agreed, or adjusting water chemistry, making drinks, etc. I think a set of peristaltic pumps gives a ton of options
It looks like there's enough space to fit three more gear-screws rotated 180degrees to run three reverse direction outputs in the offset space between the current gear-screws.