The 2020 First Robotics Competition game was titled "Infinite Recharge". It required teams to build a robot capable of launching 7in dodgeballs known as powercells into a goal 10ft off the ground. Teams were also tasked with spinning a large disk called the color wheel and lifting themselves off the ground at the end of the match.

The video on the right gives a brief overview of the game and how it is played.

We designed our robot to complete almost all components of Infinite Recharge. We could shoot multiple balls into the inner goal in quick succession, and also climb at the end of the match by hooking on to the central Shield Generator Switch. The robot could also fit under the Trench Run for faster cycle times, and could pickup power cells of of the ground with an extending intake.

The video on the left is a practice run from the end of our season when the robot was fully assembled.

By far the largest challenge presented to teams in Infinite Recharge was indexing the power cells into a single path to the shooter. The reason this was so difficult was the unique material on the outside of the balls. The outside texture was smooth but incredibly sticky. So sticky that two balls touching each other could not roll down a ramp. The squishy nature of balls also cause them to become stuck in many cracks and crevices where they did not belong. I led the prototyping of many different variations of indexing methods to consistently feed a maximum of 5 loose balls into the shooter.

The first of these prototypes used the indexing mechanism already built on the whiffle ball shooter. This was a simple brush with 4 polycord spokes. While effective with 1 or 2 balls, when multiple balls touched each other they stuck together, resisting the brush. This is because the balls were required to spin opposite the brush to roll around the outside of the track, like two gears meshing together. When more than one ball touched each other, the balls behaved like a set of three meshing gears, in that they did not move at all.

The top video on the right shows this mechanism working with off-brand, smoother balls. The brush design was no where near as effective with the required brand name dodgeballs.

The next idea we tried used the same idea but with wheels on the outside of the circular track. The idea was to combat the balls getting stuck while rotating by not requiring them to rotate at all. Instead the wheels on the outside and the brush on the inside would both spin, causing the balls to simply translate around the arc with no rotation necessary.

While this prototype was effective, it was heavy, required almost all of our wheels and collars and was simply to bulky and complex to test further. After some online research, we prototyped a completely new style of indexer.

This design featured a static wall and an angled wall of rotating wheels to direct balls towards where the shooter would be. A wooden box with belts and rollers emulated the system that would feed balls into the hooded shooter. We also added idle rollers on the corners of the the delivery chute to prevent balls from getting caught on the sharp corners of the wooden box. This required a lot of clamps and a lot of people holding various components, but this indexer design proved promising in both consistency and speed. We chose to develop this design further and used it on the final robot.

It was difficult to prototype this mechanism because of its scale and we had not yet built a bar to climb upon. Instead we drew inspiration from other teams publishing designs. I looked extensively at the design on the right at the 1:00, which featured a virtual 4 bar mechanism which deployed two hinged frames at the same time. This is the style of mechanism that we eventually used on the final robot.

For models of FRC robots which are mostly constructed out of box tubing and flat aluminum plates I tend to sketch most of the features and components in a single parent sketch and then project all the lines from this sketch for future extrusions. This means that if I need to change a portion of the design I can change it in the parent sketch at the beginning of the timeline and everything afterwards will update accordingly. This workflow has proven effective in keeping the amount of features lower for large assemblies so Fusion 360 runs smoother. However the disadvantage of this method is the sprawling complex parent sketches, which requires having a clear picture of the end result before getting started.

I chose to break up the model into a few different modules and then assemble them together in a larger file. This allowed me to do my modeling in smaller, less resource intensive files. I began with the hooded shooter, since its geometry would determine the indexer. The picture on the right shows the parent sketch of this module.

You may have noticed the complex lines in the top left corner of the shooter sketch. This is the geometry for mounting the motors which drive the wheel. I was unsure whether the belt ratio should be 3:1 or 2:1 to achieve the wheel speed we wanted on the final shooter. The motor mounts are actually two different plates that each had mounting holes for either belt ratio configuration. I was also unsure whether we would want three or four 775pro motors to achieve the desired torque, and this system allowed us to change our motors later if necessary.

In its final form on the robot this flywheel was driven by 4 775pro motors on a 2:1 ratio. This resulted in an average speed of about 6000 rpm. The flywheel has four 4in Andymark 60A Stealth Wheels in the center. These provided a good balance of grippiness and durability. On the outside of these wheels on either side are two solid aluminum flywheels to add momentum to the wheels while spinning which was essential to maintaining our rpm while firing multiple balls in rapid succession. These flywheels are 3.5in in diameter and 1in wide. On the backside of the image on the right is a square plastic housing. This holds a magnetic encoder which was used to measure and set the speed of the flywheel using a PIDF control loop.

The frame of the combined shooter and feeder is bolted onto the vertical and horizontal sections of box tubing. This frame is pocketed 0.25in thick aluminum on the bottom feeder section since it forms the majority of the structure of this module. The top shooter is pocketed 0.125in thick aluminum. The aggressive pocketing helps us reduce weight in order to meet the total limit of 125lbs. You can also see the Limelight camera mounted here. We use this camera to track the reflective targets on the field and aim the robot at the target.

The feeder pulls balls in from the hopper and raises them up to the shooter. It does this by pulling the balls with the front most roller of omni wheels, and then raising the balls with a set of polycord belts (not shown in this image) in the back. The single BAG motor accomplishes this through a pair of belts on the outside of the frame that drive a small set of gears to change the rotation direction of the front roller. The idle vertical roller prevents balls from getting stuck along the wall upon entry into the feeder

The small internal rollers highlighted in blue were added after testing the final robot and finding that balls were scraping the sides of the pocketed plates and getting stuck. These idle rollers freely spin, reducing the friction of the collision of a ball against the wall.

The agitator features two 4in Andymark 35A Stealth Wheels. These are driven by a BAG motor and spin inwards guiding balls towards the feeder mechanism. It is mounted at an angle in order to remain parallel to the angled floor of the hopper.

The virtual four-bar frame is particularly lightweight and is no where near strong enough to lift the robot on its own. Its only purpose is to deploy the hook to the bar. The lifting is done by a winch which reeled in lengths of kevlar infused cord. Two separate drums (highlighted in blue) linked by a common shaft unwind the cord as the hook is deployed, and then reel in to lift the robot. Having two winch drums ensures that the tension force is equally distributed across the climbing mechanism and robot as a whole.

The gearbox for this mechanism features an additional ratchet. At the end of every match, the field management systems cuts all power to the robot. This causes a problem for a robot held up by an electric motor. Our pneumatically actuated ratchet allows us to unwind the winch to deploy the hook, and then activate the ratchet as we climb to prevent the robot from falling back down. The small pneumatic piston pushes the pawl into a custom made gear (highlighted in blue) in order to prevent backwards rotation of the shaft.

Originally the climbing system was designed to use two hooks equally spaced apart. However, during manufacturing of the climbing mechanism, I realized that dual hooks were not ideal for our situation because the bar tilts as we climb. This means that oftentimes there is a force of gravity acting sideways on the robot. This force would be almost entirely absorbed by the lightweight virtual four-bar mechanism if we used a two hook system.

I designed a new, single hook mechanism to prevent our virtual four-bar from becoming a bent mess. This hook was mounted in line with the robot's center of gravity to minimize the amount of swaying the robot would undergo. However, the hooded shooter also utilized this space. To solve this, I created a hook that folded. In the down position the hook is blocked from rotating by back of the hooded shooter. As the climber raises, the hook is released and flips upward due to the surgical tubing wrapped around it.

While I didn't own the design of these next two subsystems like the shooter and climber, I did influence their design and they are important aspects to understanding how the robot works.

The intake is mounted on sliding rails like our prototypes, but unlike the prototypes our final design has the intake mounted at an angle. This is to allow the intake to stow within our frame perimeter and reach slightly lower than the bumper when deployed to get optimal compression on the power cells. The final intake also features a polycarbonate trough above the top roller which forms the rear wall of the hopper.

The linear sliders were custom designed by another student and printed in nylon using Multi Jet Fusion. Bearings were bolted on both sides, as well as the top and bottom. The bearings are the only contact with the aluminum box tubing, and as a result the intake glides smoothly in and out.

To interact with the color wheel we designed the wheel spinner. It's essentially just a brushless motor and gearbox with compression wheels on top. It remains folded within the robot the majority of the match, but pops up above the robot when deployed to interact with the color wheel.

The wheel spinner is deployed via a pneumatic piston, which is critical to its success. Because pneumatics are compressible, when we drive the wheel spinner into the trench it won't receive a jarring shock and break. Instead the piston compresses and increases our pressure on the wheel

The built in encoder on the brushless motor can measure the number of rotations of the color wheel, and the color sensor mounted above the wheel spinner can sense the specific color that is active on the color wheel. The video on the right shows the wheel spinner being tested before being mounted on the robot.

Most of the custom aluminum plates and gussets are outsourced to the HP Model Shop for fabrication. For almost all of these parts, I arrange a dxf layout of all the parts in the most compact form I can achieve. These are then cut on HP's waterjet. The waterjet is great for producing large amounts of parts quickly, but lacks slightly in precision. For bearing holes or threaded holes, I purposely undersize the holes so they can be drilled to size later.

Parts that see high loads and are mission critical we send to be printed in nylon on HP's Multi Jet Fusion. This material is incredibly tough, durable and flexible. On this robot we printed the winch drums, linear sliders, various piston mounts and timing belt pulleys out of MJF nylon.

Lower load parts such as spacers are printed on standard FDM printers out of PLA. For smaller parts this is ideal since the turn around time is relatively small compared to MJF.

Based on the CAD models, we can easily know how many of each part we'll need to assemble a complete robot. For gears, gearboxes, motors, electronics, wheels and most FRC specific parts we typically order from Vex Pro, West Coast Products, or Andymark. For most of our bolts and more general hardware we order from McMaster Carr.

After ordering parts, we organize parts into separate bags for each module. That way once assembly begins all the parts necessary to complete the module are in one place. The picture on the left shows a tote of parts prepared for the climber

Diagrams of custom aluminum tubing were printed out and given to the build sub-team. Each of these tubes were cut to length on our horizontal bandsaw and holes drilled on the drill press or mini mill.

The majority of our parts are held together with steel bolts or aluminum pop rivets. Rivets allow us to secure things quickly and will two thin materials together with great strength. They're also significantly lighter than using steel bolts. Bolts are used in places where the materials are thicker or standoffs attach to plates.