This is the story of my experience building “Cook-Me-Dumplings,” the 2.007 2016 champion robot. To date it is one of my most heavily thought-out projects. It was not just a robot, but rather a complex system of many parts that worked together. I sunk a great deal of time into this project but it was completely worth it.
2.007 is one of MIT’s most famous classes, started with the help of Woodie Flowers to promote the “hand” part of MIT’s mind and hand motto. The goal of the class to build a small robot which will perform various tasks on a competition field. This year’s competition was American Revolution themed: “No Innovation, Without Fabrication!” Tasks included throwing teabags off the Beaver ship, hiding cannonballs in tubes on the Old North Bridge, hanging lanterns in the Old North Church steeple, and completing the Midnight ride to Concord. One robot starts on each side of the field, and face off at the end of the semester in a single elimination tournament.
At the beginning of the class, each student is given a box of parts containing all the basic parts to build a robot. In a vein apart from many other MIT classes, each student builds their own robot.
The kit of parts includes various mechanical parts: extrusions of aluminium, plywood, acrylic, ABS, aluminum sheet, and (extra extra mild) steel sheet, as well as threaded and threaded steel and aluminum shafts. For motors the kit contains B0-P5 and B0-P6 motors, as well as various size servos. (IMHO) The motors really end up defining everything about your robot- how big it is, how much it can lift, how fast it can go. The motors end up being the limiting factor in terms of power. The 500mah 2s 35c LiPo provided in the kit can support 17.5 amps at 8.4 volts, while the BO-P5 motor has a stall current of just 3.2 amps at the rated 3 volts. Many of the 2.007 alumni I talked with had problems with these motors. More on this later……..
Down to the business of strategy analysis. Initially I looked into the strategy of completing the 150 point challenge of hanging a flag on the top of the Old North Church Steeple. This seems like a ton of points! Unbeatable, right???? 2.007 forces you to complete a series of course “milestones” in your lab notebook, initially something I scoffed at. After completing Milestone 2, it became clear that the winning strategy would actually be a ramp climber robot which simply completed the Midnight Ride task over and over throughout the two minute round, simply climbing up and down the ramp. Conservative estimates put the design at over 500 points, and thus I decided to take that route. Thus I am thankful to the milestones the course forces us to complete, even an experience old fart builder like me can learn.
Milestone Three made us draw up some design concepts. Initially I thought of making a really slim robot which would hide under the aluminum rail, and use it to gain traction. This would have been quite nice as it would be able to start from inside the starting box and just go up and down while hugging the rail.
With good control this design could have done very well, however without a mechanical link to the rail it could possibly fall off and break. This was unacceptable. The 2.007 competition is a single-elimination format, so it is exceedingly important for the robot to NEVER fail. Reliability was the single biggest design point, a key feature which drove many design decisions. With this in mind, I came up with the Concept 9 design.
The design consisted of two parts, a deployment ramp and a smaller robot which would run up and down the ramp. The ramp was needed because the robot could not start directly on the rail as it would hang outside the starting box. This design was really nice because it allowed the robot to slide onto the aluminum rail and thus be mechanically linked, similar to a rollercoaster (in fact I browsed pictures of roller coasters for design inspiration). This mechanical linking was the key of Concept 9, giving the robot much greater reliability and predictability than all previous designs.
Thus I began the CAD saga, the process of iterative design. The C9V1 design is shown below. The important parts are a set of wheels above and below the aluminum rail, as well as a set of tusks slung below the rail to keep the robot from flipping at the top of the ramp, where the aluminum rail ended.
This iterative process is one I derived from the “Rachet Up” process on my FLL/FTC team, the GearTicks. The process is as follows: 1. build thing. 2. put thing aside. 3. build another (hopefully better) thing. 4. Compare the two and use the better one. This process makes your objects become slowly more refined while saving backups in case the new version doesn’t turn out as you though it would.
It is best to cad your robots in the environment where they will operate. This robot interacts heavily with the competition field, and therefore it was necessary to cad that too:
This was a super great tool to see if the robots would actually see if the robots would fit on the competition field. It proved to be critical as the robot designs became more and more complex. I could see if robots would fit onto and around the track as well as fit inside the starting box. The C9v1 design was significantly too long to round the sharp bend at the bottom and thus C9v2 became significantly shorter, with trucks to help round the bend smoothly.
Moving to the C9v3 design:
This design changed the structure compared to C9v2. The front and back plates were replaced by plates slung underneath the rail. HDPE sliders were mounted on the rockers to constrain the robot laterally. As you can see these designs are very blocky, with only critical parts added.
As I was doing this I began to realize that there would be no way to actually drive this robot. The ramp slopes at an angle of 60 degrees mandating the use of methods other than gravity to produce normal force on the wheels, such as springs or rubber bands. Another problem was that I would need to suspend a motor on top of one of these trucks, which would be super gangly.
C9v4 fixed these issues by taking the upper wheel off of the truck and placing lower wheels on a spring.
The main problem with this design was that as the the robot left the top and the bottom of the ramp the wheels would squeeze together, and then need to be forced open again to get back on the rail. Ok, but not ideal. It was also mechanically complex as the rockers would need to be machined by hand. On the upside, the outer plate – standoff – inner plate system worked very well, and would be present on all future robots.
As problems with drive became more and more apparent I thought of a completely different way to drive the robot, attaching the powered wheel to the side of the robot, gripping the wood underneath the rail rather than the rail itself. This was just better in every way as the wooden rail didn’t end and the motor just ended up fitting better in the robot. I could also use a normal wheel provided with the kit, which was a perfect match for the B0-P5’s max power torque. As this was a significant advance I decided to move to Concept 9.1, with this new design taking the title C9.1v1.
This design was just better in so many ways. The motor sits vertically on the left side and is provided with normal force with a wheel on a spring loaded wheel on the right. The same standoffed inner plate from C9v4 are still present. Besides wheels and motor there is only one moving part. All frame pieces are waterjet or laser cut. The only pieces requiring hand machining were the eight standoffs and the Delrin wheels.
The only design change I made for C9.1v2 was changing to a stronger spring system which also would serve as a stiffener between the two sides of the frame.
I decided this one was good enough to build and submitted it as my MCM. The standoffs and wheels were machined on the Pappalardo lathes, which are very nice. Doing as little hand machining as possible greatly sped up the process of building this robot. Still, it took about three days to fully bring-up this robot.
I sorta just ignored the existence of the batteries and RX during the cad, so they got jankily taped on top.
Surprisingly, it worked.
Without the ramp-robot, it couldn’t score just yet, by my estimate put this one at about 400 points, not bad for revision 1.