Chronological Entries - 2021

YEAR SUMMARY

This year definitely showed significant progress in my skills in CAD and programming. I think after the beginning of the learning curve for me, I started to pickup certain skills and habits that allowed me to further advance my knowledge and skills. I credit that to having teachers and mentors that I was able to show my project too. I was able to get really good insight that led me in the right direction and helped me develop as a maker. I finished a course on Advanced Linear Algebra which think was the highlight of my year. It taught me so much that perfectly lined up with what I needed to know for this project!

My CAD designs not only contained more details, resulting in less problems when assembling, but overall was much more structured and thought out. I think one thing I wish I tried to do better though was more carefully document and save my code. Having just started to learn what GitHub is and how to use it , I didn't how often to utilize it as much as I should have.

1/6/21 - Working on 3D Reconstruction

I can easily do the camera on top and camera on the side deal so I can get the X, Y, Z cords of the ball. The side camera especially was hard to setup right and the top camera required a really bulky and heavy rig. One problem I was running into was that it required such a large set up with a really large FOV camera that I wasn't really liking to much.

I've wanted to go another path by using multiple angles to triangulate the position of the ball instead. This would be perfect for the final project, simple and small set up, without requiring a big over head bar and a camera really far away to capture the whole table. My goal would be to have a set up something like this:

Something with closer to 2 cameras, and a ping pong table in the middle. I think this would not only get us the correct data we need, but it would be a cool visual thing too. I know OpenCV has their own library for this type of stuff but I'm going to have to learn how.

1/12/20 - New Python Renderings

While I continue to bring over all of the code from my previous C++ code, I started a simple rendering of the Ping-Pong table and with one camera and using OpenGL. While its not completely functional just yet I plan to get the ball position next.

1/20/21 - Stereo Vision Tutorial

Here are the notes I took while I watched a basic tutorial on trigonometry and stereo vision. This is super important on how to get the 3D distance from the cameras to any point in the cameras point of view. It seems that what I want to do is some variation of this math. I will trying to understand this more and more, then figuring out how to implement in code. Super cool diagrams!

1/24/20 - Ball Translation Calculation

Here I sketched a little about how to translate the ball position to a set coordinate based on its previous position in the 3D render. After figuring out the solution, I soon understood it was a super simple solution. I realized that - how do you find the change in anything? - New Position minus Old Position.

While at first, I thought since I reached to such a beginner concept that I wasted my time and was better off just looking it up. But after looking back, I think this method of sketching out a problem and breaking it down, putting in other examples to try to figure it out, is super important for me to work out problems.

2/02/21 - Camera Issues

Playing around with the camera that last year shows it biggest flaw, exposure. To get a clean image with it, it needed to be extremely well lit which is rarely the case.

I needed a camera that did well in all light conditions, high frame-rate, and low price. I thought I could use a action camera called the Insta360 One. The camera looked super high quality, high fps and very affordable. However the webcam mode was very different. It was of questionable quality, and it recorded at 20fps, it would also have a weird zoom feature where it would zoom on faces it saw (and, no you couldn't turn this off). So I struck out with this one....

I will continue searching for a better camera for the project.

2/20/21 - New Camera

Spent the a little while saving up and playing around with GoPros / USB webcams for USB compatibility. I bought the GoPro Hero 7 and I bought the Logitech Stream Cam. The color accuracy of the GoPro out weighs the Stream Cam, however the GoPro only records at 30fps, while the Stream Cam can record up to 90fps. This is going to be good since they work over a simple USB connection can have decently color accuracy as well.

I will move forward with purchasing a couple more Stream Cams for this setup.

2/21/21 - Camera Pos Estimation

Here, I started sketching a problem that I was facing. Sketching it our and using fake examples definitely help me figure out hat to do from here.

I just sketched the situation of trying to figure out the 3D position of the camera relative to these 4 points. I still don't know how I'm going to solve this, but I have a good basis to play around with.

2/23/21 - Camera Position Problem Sketches

Here, after I played around with what I had, I started sketching a possible Idea of using sphere that have a set radius to backtrack and find the positions of the cameras. I think this will be super useful for being able to where things are relative to other things if I can calibrate the cameras every time so I am not realizing on just a physical positioning for my measurements.

4/07/21 - Continuing CAD Design!

After trying to figure out my confusion of my code dilemma, I thought it was best for me to take a break and move on to something else. Here was a sketch that I made underlying the new design principle I wanted to go after, a lightweight belt driven system. There were 2 main reasons for this change:

The Cost - The old design FAR exceed my finical budget and I thought I could do better saving money
The Requirements - I had to take a step back and realize what I really wanted this arm to do. All it had to do was move a paddle, which is extremely lightweight super fast. I didn't need a super bulky and powerful arm for this.

After realizing this, I immediately thought of cable driven or belt driven thing. They allow the source of power / motor to be located somewhere else and to transfer this motion to other joints. This means I could lightweight the arm up by moving the main actuators to the base of the arm rather than on it. Here's a quick and dirty sketch which underlines the concept for the new design. I will move to CAD when I feel ready next.

I vison this arm being a nice lightweight belt driven arm to keep the weight at the base rather than at the end effector to limit the torque required to move the arm.

4/09/21 - More Design Requirements!

With the idea created, I think its best to start with some baseline parameters and requirements before jumping into CAD. Some parameters that definitely need to be addressed are:

Overall Footprint
Overall Weight
"Arm" Weight (parts that are moving)
Joint Count (5 I feel is a good number)
"Speed" Required

4/10/21 - Arm Calculations!

Today was a super important day! I spent the day understanding the forces acting on my arm design before diving into CAD and printing parts.

Comparing estimated torques based weights I thought were appropriate along with lengths from the previous CAD model, I am able to choose what motors I think fit this criteria. I will be looking online for anyone who as made something similar to what I want, mainly keeping it 3d printable and straight-forward.

4/11/21 - Chosen Actuator!

After the calculations I did yesterday, I would like to switch from the heavy stepper motors, to lightweight high-power brushless actuators. These guys are awesome! I played a bunch with the gearbox size and motor KV to allow for slower movements but more greater torque since I don't need a lot of travel for the design I will be using.

These actuators will be powered on an ODrive! These are super fast and precise open source drivers that turn any brushless motor into basically a high powered brushless servo motor. I will go through and make sure all of the parts are compatible and begin to buy some of the components.

4/13/21 - Belt Driven Arm Concept

Here, I just found a cable driven arm that I would like to base my design off of. I plan on using belts instead of cables because I am more comfortable with how belts work and have some on hand to prototype with.

This design is of a basic 2 axis cable driven arm, but I would like to scale it up and add complexity with more axis. The design I have has 5 axis in total, which I think will be plenty.

4/17/21 - New CAD Progress!

Starting to turn the sketch into a CAD model! There are really no set dimensions yet and no parts meant to be made yet, I'm just playing around and tinkering with the concept in CAD.

I will start to used source-able parts as soon when I feel comfortable with the overall concept and where the motors will be placed. The main idea here is keeping all the weight at the base rather then out on the end effector to maintain low torques needed to move quickly.

4/19/21 - J4 Problem

As I started sourcing parts for my CAD model, I realized a problem with my original thought for the joint at the end effector.

I wanted to have a lightweight rod going down the middle with a U-Joint at the middle joint to run a joint at the end-effector so I could keep the motor at the base . However, after doing some research about how u-joint geometry works, I learned a couple things.

When a u-joint is bent the end point rotates in an Elliptical motion, not a uniform circular motion!
Looking at the data sheet, the u-joint I choose only "bends" a max of 15 degrees, no where near my requirement. Other u-joints that can bend that much are ridiculously over priced.

I will have to look into have a small motor on the end effector for that joint or probably just switch to a servo motor.

4/29/21 - Completed Table Calibration and Big Code Updates!

I have been trying super hard to discipline myself into writing detailed comments on important things and having good class descriptions to make my code as easily read-able as possible! This definitely helped me on my journey to get this process to work. While its not the exact corner position tracking that I wanted, its a really good start! I am currently able to scale the 3D render of the table based on the corners I select in the frame. I will continue to play around with this inline stereo camera setup on the table and see how well tracking works!

5/01/21 - Today I continued adding to the CAD model! I tried to design with the intent all these parts would be 3D printing. I still have a couple parts, like the side support, that are still evolving, but I plan on making these out of aluminum and CNC'ing these parts. If I don't get a chance to, I will just 3D print these parts as well.

I am also modeling the timing belts to get a sense of where I can and cant place bolts, slots, ect.

5/10/21 - Troubleshooting Stereo Vision Accuracy

In theory, this is meant to work. However the accuracy I'm getting when I target something even 3 feet away is +- 10 inches. I don't know what this is due to, but it seems the ball at those distances is such a small target, the software has an extremely hard time tracking it.

Since this stereo vision only works the cameras are inline, I cant move the cameras on either side, or pointed at angles to the table. This The accuracy drop-off of being almost 6 feet away from the ball when the ball is on the other side is a problem.

The swap to being able to place and calibrate cameras from any position seems much more involved! However, the ability to have cameras placed near the ends of the table to evenly spot the table, seems like a solution to me. I will need to do some research and work on this soon.

5/15/21 - Mid-Arm Joint

Tweaking the joint with connecting the main body tubes. This joint will have major forces in unwanted directions, so I must make sure this comes out rigid.

5/18/21 - Joint 1 Large Actuator Design

Here I found a completely DIY 3D printable actuator with an insane amount of torque for this project.

It an open source project called OpenTorque. The actuator uses a brushless motor for in source of power which is perfect to pair with an ODrive!

Here I also added the CAD model into my design, I had to redesign parts of the casing to fit in my model, and internal mounting spots because the motor that was supposed to be used was unavailable.

5/19/21 - Better CAD!

Today I worked I continued adding the remaining parts of the arm to the CAD model. I also drew up a quick to-scale Ping-Pong table to see how large the arm will be relative to the table. Looks so cool!

I will continue to work on the details and problems I see in the model so far.

I also don't know how I feel about the large belt to rotate the paddle directly. I may just want to simplify it and use a servo motor at the end.

5/20/21 - Body Tube Joint Redesign

Today, I thought of a redesign idea for the shoulder actuator. This design will allow for a counter balance of that joint. This will reduce the required torque the motors have to apply on the belts to rotate the pivot joint on the side supports.

Right now, it makes intuitive sense to me on how it will work, but I don't know if it will. It seems if we lock the side pulley to the side support and have a motor with a belt connecting this, we can get rotation from an internal source.

I think I will need to create a simple prototype to test this idea. I'm thinking some foam and rubber bands.

(picture of prototype)

5/22/21 - Prototyping the Counterbalance Concept

I spent the last day or two playing around with the idea. I started with some foam and rubber bands to see if I could recreate a mock version. After doing that It seemed to work perfectly as intended so I will continue to remodel!

5/24/21 - Counterbalance Design

Today I added the shoulder actuator redesign for counterbalance in the CAD model. Looks super clean!

I'm wondering though If I should have belts and pulleys on both side of the main body to apply an even force. I might design this to see how I feel about it.

5/26/21 - Starting with sensors!

I started adding limit switches for all the joints so I could home the arm!

5/29/21 - Finished 3D printer upgrades!

I'm using some PLA and water-soluble material to print these out.

I would like to test out some of my carbon fiber infused PETG material as well!

6/01/21 - Continuing to Tweak Code

After my little break with the code, I will spend the next bit working on a new code as I begin to 3D print all of the parts. This code must:

Implement this complex camera position estimation that I don't completely understand yet
Has organized and separate organized methods and classes for each task
Is easily readable and debug-able so I can fix problems that may occur

6/19/21 - First Steps for Assembling!

Some basic parts come and parts can be fitted for prototypes. I still need to figure out a way to trim down these tubes as well at the steel shafts. More to come!

7/10/21 - Test Arm Frame

Looks great! The only seeable problem is I may have to lengths the slot the tube fits into for a more rigid fit. I may run into problems with holes being to big or small for bearings and shafts.

7/15/21 - 3D Printing Some Parts!

These were just meant to be a sample part to test fitment and belt sizes. This gave me a good real world feel for the parts as well. These parts will probably end up being machined out of aluminum or printed out of some different material. Weight needs to be kept in mind, these things are heavy!

After testing belt sizes I realized the belt I had was too long, need to recalculate and reorder.

7/21/21 - Almost Done Printing Parts

Today, many of common parts are finished printing. Some need redesign though, some bearings holes are to small or too large. Same with bolt hole sizes. Overall though many of the parts look great!

Started 3D printing fail pile... ;(

7/22/21 - More Parts Delivered!

Aluminum plate and lazy-Susann came! These plates need to be machines for the stand support, I 3D printed a test for what I would want it to look like.

This part doesn't need to be light since it is stationary on the base, but for the other aluminum plate I might need to redesign to make it lighter since it needs to rotate quickly.

Or maybe I can laser cut out of wood!

7/25/21 - CAD Issues

I've been working on and off fixing problems that I have been encountering as I get to play around with the parts in-hand. I have made sure to:

Make things lighter weight
Added more bulk to flimsy parts
Added more main body tube support
Fixed belt lengths
Tried to add cable pathways

Also, similar to my code, I need to work a lot of Clean CAD files and extrusions etc. It makes redesigning and finding parts SOO much easier.

7/25/21 - Creating tool paths

I have about 2 years of experience working with CAM and CNC machines from the robotics club, but some of these parts require intricate CAM operations. I 3D printed most of them, I might just end up 3D printing them all (at least for version 1), I think its cool to understand the thought process of CNCs even if I do end up just 3D printing them.

IMG_1114.MOV

7/29/21 - Base Actuator!

Today I started assembling the base actuator for the arm. These helical gears are super cool, and best of all, superrr smooth!

Awesome to play around with, I will be getting the other parts for the assembly soon.

8/03/21 - More Mail!

Been getting a bunch more parts in the mail. because of this, I have been able to fit multiple bearings, shafts and belts! I still need to machine some of the shafts and plates but the parts that I can play with, I have been tweaking to work together.

I also have been keeping a log of each value that works with each part, like nice press fit bearing diameters since I kept the design to use as few different parts as possible!

8/17/21 - Been receiving lots of parts! Realized a couple problems:

Either the material or the way I am printing some of these parts leave the parts liable to break or flex to much. I think I will be redesigning more of the parts to increase rigidity of certain parts.
Shaft is really hard to get through the bearings, bearings must be lined up very well.
Carbon fiber tube mounting method will not be strong enough, I need to add more points to fasten down the tube with out cracking anything.

I will be working on resolving these issues.

8/22/21 - ODrive board!

Started working on a basic electronics board with an ODrive. I'm using this setup merely to test out my motors and gear boxes. I plan on creating a nice set of boards that sit inside the stand of the robot that houses all of the electrical components.

I also will probably swapping out the Arduino for a nice Raspberry Pi to run my software on!

8/29/21 - Image Geometry!

Here are my notes on Homogeneous coordinates. After lots of research, I realized that these, plus more, were the mathematical concepts that I was missing to understand how the Camera Matrix is estimated!

There is no way I can explain everything I have been learning about in detail, so I posted my notes here.

9/02/21 - Intro to Linear Algebra!

I'm constantly seeing this study Linear Algebra, I don't really know what it means but I'm getting the hint that it's important.

I'm going to spend the next while learning as much about this as I can. It seems like a lot of what I want to do is based on linear algebra. Like the other things taught myself, I found a course online that I will be following through for the next couple weeks.

10/04/21 - Linear Algebra!

Wow. Math is crazy. I have been learning so many incredible math concepts. I made a list of some the concepts that I'm going to have to use on my project. I thought I would document these here. Here are a bunch of different math principals that I have been researching more about how to implement these things in code. These concepts are required to be able to implement Homogeneous coordinates etc.

Above I have images of the following concepts in order respectively:

Multiplicative Inverse
RQ Factorization
Element Wise Matrix Multiplication / Division
Normalizing Vectors
Calculating Camera Matrix from corresponding points
More on Homogenous Coordinates, Calculating 3D Points after Calibration

I will talk more about these principals later on as I start to implement them into my code. Most importantly, I've spend several weeks on trying to learn about the Direct Linear Transform. I linked 2 videos that were vital for teaching me on how these principal work (there is more content on the resources used page). I also started programming a couple other of these methods into my code which I have shown that were used in other methods, some of these methods I've altered for a specific purpose like the elementwise division. I'm super excited to continue working on this, this stuff is super fun!

GENERAL FLOWCHART FOR BALL DETECTION

10/07/21 - Redesigning the Ball Detection Algorithm!

After having learned all this foundational knowledge, I will be laying out a new image detection process to more accurately detect the ball in the frames of the camera.

I have learned that laying out a flowchart for these types of situations helps getter a better understanding of how your code will be structured. So that's what I did!

Here, you can see a general flowchart for how I want my new code to be structured after learning a few different ball tracking techniques. I decided to try and stick with just grayscale detection while adding my own circularity parameter. I predict this will get me much more accurate ball locations and much less false triggers.

If necessary, I will complicate detection algorithm more, however, this is what I plan on coding up to start the new framework.

UPDATED FLOWCHART FOR BALL DETECTION

10/13/21 - Continuing the Ball Detection Algorithm

I have been busy at work getting the new algorithm to work accurately. Programming the system was easy as it only took a day or two, the hard part was tuning. After I build the algorithm I found a couple of things that I kind-of expected, but also didn't realize was going to be such a big problem.

1) When the ball is stationary it works great, however when the ball is moving fast the threshold gets a blurred image of the ball and has a hard time calculating Circularity

2) Color accuracy is very inconsistent throughout the table. They may mean grayscale will be better and actually using a white ball may show better results since it limits the variance in color since we are using grayscale.

Because of these issues, I made an updated flowchart with an attempt to fix these issues. After programming these changes into the code I saw noticeable improvements in the accuracy of the tracking. Next I plan on working on the reconstruction aspect of the code which is were the linear algebra stuff and image geometry will come into play heavily.

(video)

FLOWCHART FOR 3D RECONSTRUCTION

10/24/21 - 3D Reconstruction

It took awhile to get the logic and program to work as intended, so I laid out a flowchart for how I wanted the program to run. After that, was just a matter of learning how to program each function and in a neat and organized way. I used many of the methods and logic I learned a couple weeks ago, but used them in a much more intuitive way.

Next, I plan on taking a break from the software and continuing to work on the construction of the arm!

11/04/21 - Progress on making the real-life parts

I've been working a bunch on the actuators and how I am wiring and connecting all of them.

To sum it up, here are all the problems I have faced and have solved or yet to have solved:

Not enough thought went into cables, I'm using Ethernet cables for all of the encoders, and a 3 part cable for driving the motors.
For the small actuators, the motor shaft is too short to fit into the planetary gear box (more on that later)
For the large actuator, the gears require a little bit of force to turn once it is fully assembled, but once I added a little grease it became very smooth

I have a fully completed large actuator and one almost completed small actuator, I need to make 2 more small ones and continue with all the soldering for the connectors. Lots of soldering practice so I've been getting better with that!

11/07/21 - Solution for my motor shaft problem

Since the shaft of the motor was directly connected to the router in the motor, I couldn't switch it out.

Here I thought about using a coupler and extra shaft, I'm going to have to machine down the 5mm rod I bought and redesign the base mount slightly to compensate for the extra length.

This should lengthen the shaft just enough to fit into the coupler of the gearbox!

11/21/21 - Working on soldering connectors

Nothing crazy fun, just more progress on getting encoder cables and motor connectors all situated. My soldering skills are definitely improving from all these connectors.

12/16/21 - 3D Printed Gearbox Problems

After playing around with the motor controllers it seems the gears in my largest actuator bind up whenever there is a load put on the output.

After some experimenting I found one of the gears was deforming slightly, it seems the tolerances were way to tight for free motion. I will definitely edit the tolerances and reprint.

I will be swapping materials, and hopefully maybe machines!

12/19/21 - Almost done with the Connectors and Cables

I am trying to figure out a smart way to route these cables so they don't get caught on anything when the arm moves quickly. I am also thinking of having some common board that all connector can plug into that lead to the driver boards

12/22/21 - Further working out kinks to the CAD

The model evolving as I start to assemble some of the actuators. I also realized some problems with sourcing pulleys, so I altered them to all be 3D printable and mesh to the shaft using a metal strip.

Also tried to limit the amount of different nuts and bolts I need so I redesigned some of the mounts for parts like the main body tube holders and actuator retainers.

I will need to reprint a bunch of parts, again, so I probably should have waited on printing them all, oh well, the more you know.

12/23/21 - Closer look at new Parts

My new goal being 3D printing everything so I can rapidly prototype these parts when / if there is something wrong with them.

I have definitely gotten better at double checking if these parts will work in CAD before actually making the part, so I feel these parts are getting pretty close to fully functional!

12/24/21 - More Nuts n' Bolts!

Added the remaining nuts and bolts for all the connections that need to be made. CAD is starting to narrow down on functional design and more time is spend actually making these parts!

12/29/21 - Toolpaths!

This will probably be left alone for a little since I will just 3D print them for now, but I will have the toolpaths if I need.

I bought a PETG Tough Filament which seems super strong, but if I see problem I will pivot, however now it seems to work fine!

Page updated

Report abuse