Curivator Project

This is currently a live document, which may change as the project is still on-going; however, at the time I’ve started this document the robot is completely built  at a moment where I can pause and reflect on the whole experience with you.  This has taken 4 years to complete, but this is because it has been spread out among other activities, so it could probably could have been completed quicker.

This actual project came into existence in 2014 after returning from Worcester and seeing some of the robots at the SRR competition .  Prior to this we have experienced FIRST since 2011, and at that time most of my work was on the controls side of things, as being a software engineer seemed the best task to take on at the time.

I knew we could create autonomous controls that could solve the problems offered in the challenge, but unfortunate for us, we lacked experience in mechanical engineering.  It’s funny how we can be inspired by anyone, as a former student Keri made a statement that has stuck with me every since… She said “don’t just hope for change, be the change”.  We’ve been trying to find someone, but I decided to take this advice and start learning some of the fundamentals of Statics, Materials for Mechanics, and started learning how to use Solidworks, and so the arduous journey begins.  I had no idea at the time what I signed up for, but I started with this   drawing .  The high level idea is that it is a combination of two well-defined solutions : an excavator , and mars curiosity rover  and eventually the name “Curivator” came into existence as it is a combination of both those words.  The idea to have open bins on top was inspired from team 67 from their 2012 robot HOTbot  where they used the arm to handle multiple functions, which in essence achieves simplicity.  Curivator would then later try to adopt the same philosophy as the arm is used to sort the samples as well as become the post to hoist the camera high enough for optimal visibility.

To use the cliche that a picture is worth a thousand words, I’ve decided to make this documentary short with lots of pictures to tell the story.  I’ve kept these all in dropbox in various folders that cover various aspects and subassemblies of this project.  In addition I’ve released the cad here .  This document will provide links for each section with a brief description, and then I’ll look into providing more details in the comments section of each picture as time progresses.  I’ll also explain the overall order of things with some highlights.

The first thing that I wanted to do was to learn what I can do using 3D printing for parts.  We used 3D printing in the past for various things like gears and mounts, as well as other odds and ends for work, so I had some familiarity with them, and decided to purchase a Flashforge  printer.  I must say that learning how to 3D print is its own world of trial and error, but once you get it down, it can become an essential tool for many problems, not only for parts, but also for prototyping and making custom jigs and guides to assign in precision assembly.

So the very first thing I wished to solve was how to pick up the sample, I researched various excavators, and started to design the bucket and clasp .  I was not yet concerned about how to make the parts but more for the overall geometry, both for the bucket and then the arm.  I started with this Crayola Cad  as well as this Bucket profile sketch .  The arm went through several iterations to make sure that the range would fit .  I didn’t concern myself with the linear actuators yet, but then I did soon after as well as placement of the arm while it was driving in the stowed position .  After this and a few other designs I had help making a ½ scale model  of the bucket to test.  This bucket was made out of PLA from a coworker friend Kevin, who help get me started learning about 3D printing.  I decided to drop down  to a ¼ scale model , and wanted to complete the arm turret and payload of this.  I would worry about the drive later, so in this video  I had an overall proof of concept for the geometry, and I just threw something together in legos for the drive.  

Once I was happy with this I need to figure out how to make it full scale.  I started with the bucket as I knew I wanted this to be fully 3D printed… I looked into having it made for me but no way  was I going to pay 1700 bucks for it!  So I would have to figure out how to slice it up and then reassemble it.  I found a plugin that actually did all of this automatically, but its solution failed miserably as I learned quickly about the importance of print orientation, and the notch groove solution did not fit, and removing the support material from hundreds of little notches ran up a huge task of time that made it all abandoned.  So I finally resorted to slicing it into a notched solution using ½” #8 screws as shown here .  This took a whole month to print, and the design itself for each interlocking tooth was extremely tedious!

With the bucket and clasp  finished, I now wanted to pursue the Arm, I thought about having it 3D printed using a similar technique, but then I caught an idea of how much material I’d need even using trusses , so I wasn’t too happy about that.  I also was considering how to power them either by gears  or linear actuator .  I really liked how encapsulated the linear actuator is, as well as its gear reduction so I made a prototype  actuator to see how powerful it could be using the 393 motor.  Finally I got a prototype  of the bottom end of the boom using rejected parts from the bucket using the plugin slicer, and used a doodle pen to fuse these parts together.  This helped to demonstrate success of using a linear actuator solution, but I couldn’t make a boom printed like this, so the question came how to make the boom using less material.  I realized that having segments may weaken the overall ability of it to lift samples, so I wanted to find another solution.  Looking back at this, I probably could have used aluminum, but decided to use wood mostly because of my familiarity with working with wood, and I could set up our own tools in my friend Jeremy’s garage.  Little did I know at the time how this investment of tools would start a bunch of future woodwork projects  not only for Curivator but other projects as well.  The idea was to use the wood to provide overall cohesive strength among the printed segments .  Where the segments were designed to use the least amount of material, but keep strength around the jointed parts , and also support  for holding the linear actuator for the boom .  At the time of the prints we used ABS, and it was difficult printing some of these parts because of warping, and in some cases we had to glue  down the corners.  For the big arm  we used the printed segments to define the cutting profile for the wood, while for the boom  we simply measured it out on a draft board, both techniques worked fine, but the big arm would need the other method because it was too large to be drafted in one setting.  Once the boom and bigarm were made we were able to assemble them as shown here .  I had not yet determined what to use for a linear actuator, and even though I somewhat knew the actuator would need to be pretty strong… I had no idea so I used string and pulled by hand to get an idea, I was considering to use a printed actuator, but after this test  I regretfully chose to use the Dart  and so now it looks like this .  I say regretfully because of its weight, while it’s the right size it was made extra strong and weighed more to be able to handle heavier loads, so it’s not quite the ideal solution.  I just had to move forward knowing I have more weight to manage than I cared for, but there was one benefit that I didn’t realize at the time, and that was that I could use the back end to mount  the Zed camera.

With the arm completed, I needed an arm mount  and for this I found some great lazy susan parts to use from mcmaster as a test I had my daughter Athena  stand on end while I rotate the arm around.  In my preliminary design , I started with wood, but later decided to switch over to versaframe , as it was lighter and took less space.  This became more apparent when I became exposed to the steel corner braces they offer at Lowes.  I also needed to design a way to power this, Jeremy helped me with the idea of a planetary design , where I could achieve great gear reduction using less space.  This came to pass as I bought some of the Vex EDR kits for other projects and really wanted to make use of their larger gears as well as using the 393 motors.  I came up with this .  It took a bit of tweaking on the actual printed measurements.  On the first attempt , it was too tight noisy as the gears meshed, and then fixed this using thicker gears .  For the outer ring, I made thin prototype rings  to get the fitting to work.  At first I manipulated the number of teeth , but found this was not the right thing even though the test results  were successful.  So instead I manipulated the pitch to get the ideal fit  as the gears mesh in the centered position from that.  To put this all together I needed the turret platform .  At first the turret platform was the entire payload platform , but later decided to separate it out via hinges  to access the WEB  (warm electronics box).  Once this was made I used the director's chair  to hold the arm, and I’d keep it here for over a year until I made a mount  for it and then some testing braces to test on the desk area  again.  While testing, I used an older controller the cRIO as shown here, luckily all the code is written in c++ 99 so that it remain compatible for this system.

Meanwhile, we needed to work on the drive  which consists of a Rocker and Bogie  design similar to Curiosity and the WEB  (warm electronics box).  For now I’m going to give a brief overview of each revision, but later come back and provide more detail.  At first we wanted skid steering as this  was an attempt to clamp on the rocker and bogie to the pods , but as I got to thinking about the traction of these wheels I really wanted to avoid skid steering and pursued Swerve Drive , which like on the curiosity can pivot on each corner wheel to turn smoothly without skid.  So without swerve is version 1, Version 2  is with swerve, and even without load I noticed a lot of deflection in the rocker frames, and the aft rocker frame had torsion moments causing the bogie wheels to bow out.  I tried to fix this in Version 2.5  by making the bogie pivot smaller, and while this showed signs of success with the 2.5 1/4 scale model  it still wasn’t enough support.  So then onto Version 3 , which indeed put the support right below it, the problem with this design is that the wheel was not powered and would cause the outer center wheel to lose traction… in one test on of the NeverRest 20 motors became defected (too much load, on what later was defined as a defected design) I had to replace the motor with Orbital 20  replacement, so I then moved on to a Version 3.5  solution, which allows the aft rocker to move on its own axis.  Traction problem solved, but now it would flip outward and collapse on itself I spotted this on the  3.5 1/4 scale model , but once again I didn’t give it any merit as my short burst tests didn’t show this happening… it wasn’t until I tried to demo the robot for the boy scouts with a significant distance forward that I noticed this, which only happened with the payload platform… that caused more deflection making the center wheels skew outward, so this problem is address in the Version 4  solution.  This design requires several revisions within itself to ensure the joints where enforced, and I had great success on initial drive tests ; however, once we completed the payload  (which is estimated around 25 pounds) it caused too much deflection rotation in the bogie pivot area as shown here , where the aft rocker shaft gives way… the inner wheel jets forward enough to cause the outer center wheel to skew outward… and over time it can even bow out like in version 2.  Without the load… these problems go away, so this solution may still hold true, but it bothers me that it can support much weight, so now onto Version 5 , hopefully  it may work, so this is where I’m currently at on the drive.

        The other thing to touch on is the Camera Mounts , which were pretty easy given the load is light, at one time we wanted to have a mast  to mount the primary camera, but after thinking about this it made since to make use of the big arm itself… especially the casing of the dart, to create a mount over it.  This also allowed the front of the robot to be clear and have a clear driving configuration.  So with that we have the robot fully built , with many Video Demos of various tests to get to that point, but it is still an ongoing process.  Stay tuned to this document as I’ll keep updating it as things get completed.  For now all the links within this blog should reach all the major folders so these will also be updated.  In particular the video demos folder as I hope to have some cool autonomous tests, but all the mechanical problems need to be solved before that happens.