วีดีโอ

I would recommend a capstan pulley if you are trying to increase precision

30 degrees as angle inwards seems a bit too steep for the jacobian. Is that a realistic angle? You know the minimal throw height for juggling 3 balls, that throw height gives you the maximum throw angle inwards where it lands in reach of the platform. The more realistic you pick this angle, the better the optimization of the platform

you should teach for a living

Abstract This video shows a conversation between two robotics enthusiasts, focusing on the development and challenges of a juggling robot named Juggabot. The conversation covers both positive advancements and setbacks encountered during the project. Key successes include the implementation of a new hand design with a smooth spool mechanism and improved cable management for unobstructed movement. However, technical difficulties arose with the CAN bus communication system, leading to a fried transceiver and unexpected behavior from the Odrive motor controller. The discussion delves into control system strategies for achieving accurate and reliable throw-catch trajectories. Options such as the Ruckig trajectory generator and the possibility of direct current control are explored, considering the trade-offs between complexity and performance. The conversation concludes with a demonstration of a functional throw-catch rig and a shared optimism for overcoming the remaining challenges to achieve successful robotic juggling. Good News 00:01:30: Discovered that the Dyneema string used in the robot stretches over time, but creep testing suggests it will reach a peak after a while. Plans to pre-stretch the strings to prevent further stretching in the robot. 00:04:30: Received a free Odrive regen clamp from Oscar at Odrive, which will significantly improve power management in the robot. 00:07:00: Juggabot is looking great with the new central cable setup for the hand Odrive, allowing for smooth and unobstructed movement. 00:09:00: Successfully swapped the hand over to John's design, featuring a smooth and solid spool mechanism. Bad News 00:11:00: Fried another CAN transceiver due to impatience with the setup. Experienced weird symptoms where the Teensy received messages but the Odrive wouldn't respond, requiring further investigation. 00:19:30: Learning that control systems are difficult, but grateful for John Benno's expertise. Experiencing challenges integrating the Ruckig trajectory generator into the CAN code. 00:25:30: Had a frustrating day trying to get the Ruckig generator working, but John has had more success with his hand mechanism and throw-catch testing. Meeting with John Benno 00:26:00: Successfully connected with John Benno after technical difficulties. 00:27:30: John demonstrated his functional throw-catch rig, showcasing impressive throw height and catch accuracy. 00:45:00: Discussed control system strategies, including the use of position, velocity, and torque feedforward components, as well as the possibility of simplifying the system by going directly to current control. 00:50:00: John shared his script for generating throw-catch trajectories using the Ruckig generator. 00:50:30: Discussed the importance of motor temperature management and the limitations of the current motor size. i used whisper and gemini 1.5 pro

I think you won't reduce the creep to zero with the pre stretching but substantially reduce it. The creeping can be seen in three phases. In the first phase it yields a few percent in a couple minutes or hours. Then in the second phase it will have a constant stretch rate over a couple of years and finally in the third phase it will quickly increase in the stretch rate and break. The following link gives a good overview! www.samsonrope.com/docs/default-source/technical-bulletins/tb_understanding-creep_mar2012_web.pdf?sfvrsn=1e71fc4e_2

Just commenting because my name is also Harrison and is Harrisons gotta stick together.

18:30 The ODrive problem sounds like the drive computer is getting stuck, and when you move it by hand it'll produce some internal event that makes it unstick. Basically a bug in the ODrive, so dunno howi it could be fixed, besides upgrading the firmware.

Unfortunately the last few minutes were chopped off due to the huge lag we were experiencing today! Very frustrating, but thankfully the bulk of Jon's and my conversation was preserved.

If I buy 4 of the same linear actuators and want to use one switch to activate them all simultaneously how would I go about wiring them together?

For the string, you could use fishing wire which you can buy on a reel and has a plastic coating. Different wire gauges will obviously give you different stiffnesses.

I understand if this is a hobby or if you are making this as a future product, otherwise it's always cheaper to buy this stuff so long as you value your time at market rate. When infirst started my automation company I designed everything, but as time went by I came to realize it's cheaper, faster and of higher quality if I buy a COTS item rather than Ddesign.

Can you link to the space mouse that you use?

For the belt option not running smoothly. You think it might be in relation to the cantelaliver set up you got going on? If the linear carriage has lost a few balls it would affect its ability to resist a moment. So another linear rail on the other side might fix the problem. Just a hunch

*Summary with Timestamps:* *Hand Actuator Design: (Discussed throughout the transcript)* - Current design: Utilizes a linear guide rail with a string-driven mechanism for actuation. - Strengths: Robust, efficient, and easily adaptable to different platforms. - Potential improvements: - Reduce weight of the linear guide rail, potentially using a 9mm rail instead of a 12mm rail. (Mentioned around 17:30 and 45:00) - Investigate alternative materials like carbon fiber for weight reduction, possibly for the top ring or linear rail. (Mentioned around 46:00 and 53:00) - Optimize spool diameter for efficiency and inertia considerations, experimenting with different sizes for testing. (Mentioned around 38:30 and 40:00) - Refine string termination method for increased security and ease of replacement, potentially using a splicing technique or a more secure knotting approach inspired by John's design. (Discussed extensively between 24:00 and 33:00) *String Routing: (Discussed throughout the transcript)* - Current design: Employs a tensioning pulley system to guide the string around the hand actuator, with potential use of belt idlers for additional string movement accommodation. - Strengths: Provides efficient and secure string routing. - Potential improvements: - Investigate the use of belt idlers to accommodate string movement during operation, ensuring adequate clearance to prevent string interference with other components. (Mentioned around 11:30 and 59:00) - Refine the placement and routing of the string to ensure it does not interfere with structural elements, particularly the support rib for the motor mount. (Observed and addressed around 1:45:00) *Platform Structure: (Discussed throughout the transcript)* - Current design: Utilizes aluminum extrusion for a rigid frame with modular components. - Strengths: Provides a strong and adaptable platform for the hand actuator and other components. - Potential improvements: - Explore the use of carbon fiber tubes for weight reduction while maintaining stiffness, potentially for the main structural elements. (Mentioned around 46:30 and 53:00) - Integrate mounting points for sensors like laser gates and IMUs, taking advantage of the modularity of the aluminum extrusion. (Mentioned around 1:12:00) - Refine the design to ensure the hand actuator is centered within the platform ring, achieved by recessing the aluminum extrusion into the ring by 20mm and adjusting the hand design accordingly. (Discussed around 1:00:00 and 1:05:00) *Control Strategies: (Discussed around **42:00** and **1:18:00**)* - Current approach: Employs open-loop control with trajectory generation, using trapezoidal profiles with higher deceleration for faster acceleration. - Future considerations: - Implement closed-loop control using sensors like laser gates and IMUs for improved accuracy and adaptability, particularly for throw height and catching. - Develop a more sophisticated trajectory generator to optimize throw height and velocity while respecting actuator limitations, potentially using spline-based methods or other advanced techniques. *Software and Tools: (Mentioned throughout the transcript)* - CAD software: SolidWorks is used for design and modeling, with potential future transition to Onshape for collaborative capabilities. (Mentioned around 1:25:00) - Version control: The project utilizes a versioning system for tracking design iterations, with the current platform at version 6. (Mentioned around 1:29:00) - Collaboration: The individuals are actively collaborating and sharing design files to facilitate development, with John providing valuable insights and suggestions for improvement. *Additional Notes: (Mentioned throughout the transcript)* - The importance of a space mouse for efficient CAD work is emphasized, particularly the SpaceMouse Pro model. (Discussed between 47:00 and 50:00) - The challenges of naming conventions for parts and files are discussed, with inconsistencies acknowledged. (Mentioned around 17:00) - The value of online resources like the SOLIDWORKS Expo for learning and improvement is highlighted. (Mentioned around 52:00) - The transcript also touches on personal experiences and preferences regarding CAD software, design philosophies, and workflow. i used whisper and gemini 1.5 pro to create a transcript and summarize it

For belt attachment, just make a belt-shaped slit (i.e., smooth on one side and zigzag on the other, to match the zigzag of the belt), where you can insert the belt sideways.

best routing solution put the string through the tube aka carbon fiber tube. Period.

I haven't read thru all the comments yet but, what about a mini ball chain instead of synchro-mesh?

Sir I was the one of the people asking for an update video (asked in the last video). Thanks for the response there and now here to the other person asking. If you're feeling you're close, yeah I understand it's better to focus on finishing it. I decided I took some time to watch the livestream and want to comment on the capstan. I agree that it is engineering wise a very neat solution. But in regards to inprinting the syncromesh into the capstan, I think that will introduce a problem. If you look at the threaded capstan, the wire moves up and down a lot. The same will happen when imprinting the syncromesh because that can only be done in a spiral I think? And that's because the cyncromesh cable isn't symetric along it's axis Something you could try is to use dog tag chains/ball chains, which are symetric along their axis. Then you can simple design long grooves along the axis of the capstan. This is probably easier to print too.

Just wrap around the string multiple times. Friction for string around a drum increases exponentially with contact angle. This will fix your issue with this design

You could use rollers with springs so that they can travel sideways. So that you dont have to change the angle of the tubes. This would also allow sidemovement for the hand as you mentiont but it would prevent that the hand gets tilted. So that one side is higher than the other. I hope you understand what I mean. Im following the project for quiet a lont time and really love it! So keep going!

This reminds me of a mechanism I saw from a company called rise robotics. They make different style actuators now, but 10 years ago their first one was called a "differential conical drive." It's similar but overcomes the issue with the capstan slipping.

Curious, what's up with the robe? Pulling a Hugh Hefner look? Haha PS Keep working on it, it's great.

Your spool problem seems like a great option to try SLA. Resin printers are cheap these days and can make fantastically small features, could help you in other future areas where you need very good tolerances.

8:42 For measuring precision, I bet you could mount the actuator to a pivoting mirror and reflect a laser off the mirror. Then, depending on how the (measurable) laser point moves with your (incredibly small) movement of the actuator, you can use trigonometry to determine your actuator's precision.

Brilliant

To keep the motor more cool, you could add a small fan to it that is driven by the motor. This way the motor would cool itself when operating

Sir could you probably make another project update video. I don't have the time to follow the weekly livestream so I'm a bit out of the loop. Probably also the case for other people. Like what is the grand scheme of the project, what do you want to accomplish? I subscribed to your channel back at your last update video about the linear actuator as I found that very interesting. Is the design of the actuator now finished? What works well and what doesn't? Anyway, keep up the good work!

I'd love to see some normal videos again, I unfortunately don't really have the time to watch 1-4 hour livestreams

When a motor blows up - the windings have heated up enough for the enamel to smoke off, leaving bare wires that short against each other and stop the motor from being functional (it'll only be able to vibrate, if able to move at all). Hopefully the ODrive can pick up when a motor has failed short like so

You're not going to re-charge directly into batteries with a system like this, it's too much too fast, it would have to go into caps.

Might be tangential to what you’re making, but skyentific has a cool pulley mechanism. th-cam.com/video/utDagouxM5U/w-d-xo.htmlsi=Fpv-Oywu3FJOgfkG

Suggestion, split the updates and bulk stream into two uploads, I enjoy watching the updates but a 4 hour vid is too intimidating to click. Keep up the good work!

5:00 Yeah, that's definitely not new, and also very obvious. 6:54 I don't quite believe that you find it counterintuitive if you think about it. Imagine having just a loop around the wheel a couple times. Then you could rotate it infinitely much, but there's a risk of slippage (although it's less than if you only had a single loop). To fix that you attach one point of the string to the wheel. Now there's no slippage, but you can rotate it only as many revolutions as you have loops. Also, no need to fiddle with it when making the loops, but simply loop only one of the ends. That way it's also easier to align the end point. So exactly how many revolutions will it do? Just count the loops. If, say, ⅔ of the wheel has 4 parallel lines and ⅓ of the wheel has 5 parallel lines, then the range is 4⅓ revolutions.

After the stream I closed SW and after re-opening, I was able to right click on broken features again! So strange! Must've been something to do with me streaming 🤔

What is the cost behind this mechanism??

Rotate the motor 90 degrees and get rid of five pulleys AND A WHOLE LOT OF FRICTION! 😁

As far as drone motors go, you could use bigger one like a 2812 motor. It will handle the heat better and proven to run on 12s that some bigger cinelifter drones use

C'mon, molex are the worst! The stupid pins flail around wildly, they grip way too hard, and the connectors don't have big enough ears for disconnecting them. The least horrible ones have a kind of flaps that you press to disconnect, but they're not working very well either.

For small soldering jobs you should use a heat plate

If you isolate the system ground from the chassis ground, using an oscilloscope on the circuit will require a differential probe, because otherwise a "normal" scope's probe ground is connected to its chassis ground, and if you connect the probe ground to the system ground, you'll effectively be connecting system ground to the chassis ground through the oscilloscope's probe which might cause some erratic behavior. That said, I don't know if you need to put an oscope on the circuit ever. Just a consideration.

A push-bike handbrake cable may be a good solution for your string problem. These are bowden cables and use stainless steel cables. Theses aren't flexible but you could use your string inside the outer sheath.

It goes without saying that I would've made a radically different design, which is not necessarily a good ting, but given the ground rules I do have a couple of observations/suggestions. Have you considered that these rods, wherever you get them, are most certainly not particularly precise? Have you considered making a lathe kind of setup with abrasive or similar, to ensure they are actually round and the right dimensions? As for the bearings, have you considered that it might be advantageous to have 3 layers, each with two opposing bearing? Really, you could do as many as you like, and could even get away with not having opposing bearings, lowering friction, leading to less wear and tear, Or how about tilting the bearings a bit, forcing a slight rotation of the rod, presumably leading to more even wear and tear? Or how about hexagonal rods?

I didn't watch your whole video, but maybe use a square aluminum tube actuator shaft to solve the wobble problem? This would help with bearing sleeves as well. Also, could have channels routered into the tube to secure Bearings in place better, too? IDK, just a thought, nice work, man

You mentioned removing the CAN transceiver chips from one board to acquire replacements. I'd recommend against this as removing them is a big hassle without a rework station and even with one or sucks. If you order naked chips you can destructively remove the broken ones and save a ton of headache

*Abstract* This video details the progress and challenges encountered in Project DeepBlue JuggleBot, a juggling robot project. The main focus is on the catching mechanism, with recent tests revealing excessive inertia causing the ball to bounce off the hand. The video explores potential solutions, including using smaller motors and a bidirectional drive system, and discusses the possibility of incorporating ball tracking and D3O impact-absorbing material in future iterations. *Summary:* *Housekeeping (**0:00**)* * A simple fix for overly tight XT60 connectors was discovered. *What Went Wrong? (**1:53**)* * A ground connection issue caused damage to several CAN transceivers. * The extent of the damage is still being assessed. * ODrive Pro might be more robust against this type of issue. *What's the Damage? (**4:40**)* * At least one ODrive and the TNC transceiver board are fried. * Replacing the CAN transceiver chip on the ODrives might be possible. * Thankfully, the Jetson computer is undamaged. *How To Stop this Happening Again? (**7:00**)* * Using bus bars and soldering directly onto the ODrives are potential solutions for preventing loose connections. *ODrive (and motor) Delivery! (**12:24**)* * New ODrive Pros and small, powerful drone motors have arrived. * A decision needs to be made whether to switch to the new ODrives or attempt to repair the old ones. * Gearing the new motors down might be necessary, but could increase inertia. *A Question About BLDC Motors (**15:14**)* * Can the new motors be safely powered by a 45V power supply despite being rated for 22V? * Overpowering might increase speed and heat, but the ODrive can regulate speed. *Progress on Catching (**17:03**)* * Catching tests have been conducted with a compliant hand design. * The hand throws the ball very high, potentially hitting the roof. * An unexpected problem arose where the strings pop off the pulley during fast movements. * A solution involving a tapered ceiling above the pulley might be implemented. *Thoughts on Inertia (**33:01**)* * The current setup has too much inertia, causing the ball to bounce off the hand. * A test rig will be built to investigate the relationship between ball and hand inertia. * Smaller motors with low inertia are promising, but gearing them down could negate this advantage. *Brief Tangent on Ball Tracking (**36:48**)* * Ball tracking is planned for future iterations, potentially using a 3D camera or a laser gate. * However, open-loop control might be sufficient if the robot is precise enough. *Returning to Inertia (**39:02**)* * Gearing down motors increases their effective inertia. * A test rig with different gear ratios will be used to assess the impact on catching performance. *BLDC Thoughts (**42:49**)* * BLDC motors are not significantly affected by voltage, but heat could be an issue. * The hand motor's duty cycle will be low, potentially mitigating heat concerns. *Pre-empting the Ball While Catching (**45:30**)* * Pre-emptively moving the hand to receive the ball could reduce the required motor force. * This has been tested, but the current setup's inertia still causes bouncing. *Thoughts on the Constant Force Springs (**48:08**)* * The constant force springs might be replaced with a bidirectional drive system due to reliability issues. * This would require careful string routing and potentially another fuse to prevent the hand from hitting the base. *D3O (**53:02**)* * D3O, an impact-absorbing material, might be used to line the hand and reduce bouncing. *No Stream Next Week (**53:59**)* * There will be no livestream next week due to the Easter holiday. disclaimer: i used gemini 1.5 pro to summarize the youtube transcript.

The density of your ball is so different from the ball bearing they dropped on the D30. Doubt it will provide the solution you want. I think, given all the articulation you've created, it's mostly software to track the ball and move the catching mechanism to gently receive it. This could also serve as a sort of closed loop system where unseen variables altering the ball's speed/position/trajectory would be addressed by tracking the actual ball

41:10 ish - yeah this is a thing I also don't love about the larger (of the smaller) drone motors like that (yes! they get SO much smaller!)- but because the propellers are so light, they don't have lots of rotational inertia, so the likelihood of them slipping is very low. Moreover - plenty of motors with that propeller mounting standard do have teeth cut into the base where the propeller sits, so when you cinch down with a locknut, it digs into the prop some and provides some mechanical keying

Just realised for around 20:00 it's relevant to note that I'm using UHMWPE (dyneema) thread, which (AFAIK) doesn't stretch *too* much, but obviously seems to at least a little...

You might get finer control of the breaking strength by using a smaller nozzle, like 0.2mm, which would double the number lines to needed to make the same diameter.

For quick power connect/disconnect, you might consider barrel connectors. There are wireable female and male versions of these.