Blog

Build Blog Day 6 (1/15/18)

Gearbox
Task: Design drivebase gearbox

  • We found out today that the gearbox will have to be redesigned. Our original plan consisted of a triangle of mini-CIMs but this design was too tall and would not allow the tunnel to lay flat and exhaust cubes to the exchange. Our new design still uses 3 mini-CIMs but they are arranged flat so that they can be lower than the bumper rails.

Tunnel
Task: Design tunnel with timing belts to friction drive the cube across on all four sides

  • For the tunnel, we worked on prototyping a mechanism that used timing belts to move the cube.. We made a four bar that allowed us to test the cube in both orientations (11 and 13 in width). Initially, we had trouble with the prototype as there was too much friction between the wood the cube rested on and not enough tension in the belt. We redesigned the prototype and made changes including making the four bar structure more secure, tensioning the timing belt, and installing plastic rails for the cube to slide on. These changes greatly increased the effectiveness of the prototype, and pushed the cube quickly through the tunnel. We now are designing a variation of this but using rollers instead of timing belts, to see which one works better.

Programming
Task: RPLIDAR Driver

  • After ensuring that the RPLIDAR works, we discussed several options, and have started to implement two of them. For our first method, we were able to use existing examples of code, from the RPLIDAR’s Software Development Kit (SDK), and write our own program that gets data and a timestamp for that data from the RPLIDAR, and send that data to the robot. Because the SDK is for C++, we would probably have to run this program on a co-processor on the roboRIO that communicates with the rest of the robot code over User Datagram Protocol (UDP). Our other option, which is more preferable, is to use the Java Native Interface (JNI). JNI allows the Java Virtual Machine, which runs Java code, to call functions written in other languages, which will allow us to use the SDK, even though it is written in C++. We will continue work on both of these methods, and choose whichever will work better for the robot.

Task: RPLIDAR Visualizer

  • We scaled down points to the size of the field, but this messed with the zooming and panning functions so we’ll have to fix this. Here is a sample image of the current functionality of this interface:

Build Blog Day 5 (1/13/18)

Drivebase

Task: Model drivebase moving up platform ramp with 1/8" center drop and minimal space from outside bumper supports

  • To ensure we have a stable robot when our CG raises with our elevator, we decided to extend our wheelbase as far out as possible. By doing so though, we are more susceptible to high centering on the platform ramp. To ensure we do not do so, we revisited the sketch we used to model the drivebase motion and adjusted the parameters to space the wheels as far out as possible and with a center drop. We also decided to constrain the lowest height of the bearing holes by having the center 16 tooth sprocket with chain OD be tangent to the baseplate. We will account for the chain run by having a cutout in our base plate.

Task: Finalize drive gearbox

  • We finalized the design of the drive gearbox and placed an order for the necessary hardware required (gears, pneumatics, screws, etc.) We hope to begin assembly of the drive gearbox this week.

Task: Sketch electronic placement

  • Task: Sketch electronic placement
    We sketched the electronics placement on the baseplate. We plan on having the battery lay flat as in 2017 rather than standing up as in our 2016 robot. This way, we can lower the carriage/tunnel as far down as possible which means the cube does not need to move vertically as much from the ground to the handling mechanism.

Intake

Task: Prototype Intake

  • We assembled an intake with adjustable wheel placement and pivotable arms/outer wheels, and mounted it to the 2017 drivebase. We controlled each wheel with its own independent 775pro geared with a 5:1 reduction —- by doing so, we were able to test asymmetry and easily adjust certain speeds of certain wheels. This intake prototype worked well in quickly drawing in the cube and orienting it from a diamond to a square shape. We also saw that by increasing the pressure in the pistons controlling the outer arm pivot, the intake itself became a lot stiffer and was able to better control the cube. We quickly implemented a bottom kicker roller and saw that it worked well because the intake itself was able to provide enough force to push the cube up and above the kicker roller. Moving forward, we will design a prototype that combines this intake with a kicker roller and mount it on a test drivebase.

Task: Prototype ramp intake

  • We design and tested if a kicker roller + a ramp with wheels on either side guiding the cube up would work in transitioning the cube from the ground to the tunnel. While testing, we realized that the wheels on either side did not grip well to the cube (we used colsons) and the transition from the kicker roller to the ramp and wheels was not as straightforward as expected. When the cube goes over the kicker roller, its motion is not very controllable especially when that is the only roller it is contacting. We added a vertical constraint (a piece of 2×4) to stop the cube from being kicked too high, however it deflected the cube from the center of the intake and away from the ramp. Watch this video to see the prototype in action.

Miscellaneous CAD

Task: Cartoon CAD and system integration of robot

  • Today was mostly filled with discussion regarding how best to integrate the various systems of the robot and how to constraint the placement of those systems. In terms of the elevator, we will be moving the outer bars (2×1 box tubing) to sit on top of the frame rails. The outside face of the intermediate stage will come in 0.138" (same clearance used in 2011) from the inside face of the outer bars, and then the carriage tunnel will be another 0.138" in from the inside face of the intermediate stage. This leaves us with around 17" for the width of the tunnel which we think will be enough to handle any orientation of the cube. We also investigated the elevator placement with regards to climbing. To make climbing especially easy on the drivers, we plan on driving straight up onto the ramp and having our bumpers flush against the scale wall and our elevator bar flush against the 2×2 box hanger box tubing. With those constraints, this means our elevator sits 8" from the outside face of our bumper which also means that given we want to spit cubes out on the back side of the tunnel (the side opposite the intake), we need to spit them 8"+. 8"+ is not ideal especially if we want to strategically place a cube, so we need to further investigate how best to optimize elevator placement. We also did line contact calculations on the bearings contacting the intermediate stage from the outside bar, and given that we are hanging 3 robots, then the bearings as well as the aluminum tubing will endure more load than their static load capacity and yield strength, respectively, can handle, To solve this, we tossed around ideas of alleviating a portion of the load from the bearings with a tensile member extending from the top of the intermediate stage to the CG of our robot (the tension of the cable acts opposite to the torque of the robots). Lastly, we decided that to ensure we have a smooth transition from the intake to the tunnel, we will have the intake fixed to the tunnel and not pivot, but will have the entire tunnel + intake system pivot such that it can stow within the frame perimeter, intake horizontally, and place cubes on stacks on the scale.

Programming

Task: RPLIDAR Driver

  • We tested the driver that was already written, but nothing happened. After rewriting some of the code, we were able to send a request to the RPLidar and received some sort of "ok" response, but still, the RPLidar did not start scanning.

Task: RPLIDAR Data Visualizer

  • We worked on some more modifications to the visualizer. We were able to cache all the data sent to the visualizer in an array, so that we could view data with specific timestamps. We also worked on being able to zoom in and pan around the image of the field, so that it is easier to see, but there are still some bugs with this.

Task: Intake Prototype

  • For testing purposes, we remapped the control board so we could toggle each of the four motors on the intake prototype with four separate buttons on the driver station. We also added some buttons to change the speed of the motors on the intake. In addition, we read the drive code, but since last year's programming bot did not have a working compressor for the dog shifters on the drivebase, the robot could not drive.

Build Blog Day 4 (1/12/18)

Drivebase

Task: Test different wheel arrangements between colsons and omnis

  • We decided to switch our initial idea of using 6 colsons to 2 colsons and 4 omnis on the drivebase. We also decreased the drivebase width by 0.6 inches to accommodate for the entire width of the colsons. In addition, we added custom wheel shafts into the Drivebase CAD. We also added the gearboxes into the drivebase CAD. This will require us to change the placement of the electronic components.


Gearbox

Task: Design drivebase gearbox

  • We started by resolving a few minor issues with packaging Cims around the shifter cylinder. We also realized that we needed more space for components in our gearbox, so we decided to increase the size of the gears between the Cims to fit all necessary components. After fixing these issues, we now have a working gearbox design. At our next build, we plan to finalize the assembly and detail the Gearbox CAD.


Intake

Task: Design intake prototype with two horizontal rollers to pull in and kick up cube

  • We began build by testing the 1st simbotics style intake prototype with 2 sets of wheels. After our testing, our results confirmed that asymmetry is the best way to go forward. Next, we assembled a new intake prototype with a horizontal roller and a kicker roller placed slightly behind the horizontal roller. When testing this prototype, we noticed that the cube would be lifted off the ground and get stuck between the rollers. We plan to keep testing this prototype since the kicker roller showed some promise in getting the cube over the frame rail. Look at Video 1 and Video 2 to see this new intake prototype in action.

  • Lastly, we laser cut version 2 of the simbotics intake, which will be assembled during our next build.


Tunnel

Task: Design tunnel with timing belts to drive the cube across on all four sides

  • The original plan for the tunnel was to have a square tunnel 13”x13” that would have power belts on each side. This way, no matter the orientation of the cube, it would always be in contact with at least 3 sides.


Programming

Task: Test the lidar driver with the roboRIO

  • We worked on the driver, which helps us get data from the RP LIDAR sensor. The sensor gives the distance in mm and the angle in degrees of a detected object.

Task: Hook up the visualizer to the robot

  • We also worked on the visualizer by plotting dummy points on a picture of the field. We encountered a few bugs, but fixed them all. Look below for a picture of the visualizer.

Task: Setup and testing

  • Limelight is a camera specifically made for FRC, which makes it easier to use, so we have been experimenting with it. We tuned the vision pipeline specifically for the power cube. In this process, we realized that we had to re-tune the vision pipeline in different lightings, so at each competition we would have to re-tune the pipeline if we were to use the Limelight. The Vision Pipeline is a set of consecutive image processing functions that work to isolate and find specific objects.

Build Blog Days 2 and 3 (1/8/18 – 1/10/18)

Drivebase

Task: Test drivebase with modelled dimensions (28"x33") 1.15” center drop; from top of frame rail and placed 2.5” from the front and back of the drivebase frame high centering on the ramp at various angles.

  • When approaching the ramp perpendicularly, the robot clears the ramp and does not high center. However, when it approaches at angle towards the corner of the platform where the front and side ramps meet, it high centers. Also, when the robot is travelling up the ramp with the cable protector below it, The back of the drivebase frame contacts the cable protector and high centers on it. The drivebase modelled with these dimensions in CAD accounts for a 16 tooth sprocket with chain OD and a 0.1" clearance between the sprocket-chain OD and the baseplate. In addition, the 4" colson wheels used are actually 3.9" wheels. If we were to bring the sprocket lower to the baseplate and use fully 4" wheels, there is a possibility it would not high center. We need to investigate the lowest we can go with the sprocket-chain OD and test with new 4" wheels. Additionally, these results may change our driving strategy where we will only attempt to go up the ramp perpendicularly. See these videos to see the drivebase in action.

Task: CAD drivebase weldment and begin work on drivebase assembly

  • We designed the drivebase frame rails such that the wheel placement and center drop can be easily modified as they are yet to be finalized. We went with our traditional West Coast Drive using 4” colson wheels. Assuming we will want to carry our partners while climbing, we need to beef up our drivebase, whether it be using thicker box tubing or by adding cross members. We will perform a bending calculation on different sizes of box tubing to determine what is the best size to use. Regarding bumper mounts, we currently see to ways to approach the task. The first way is a 2014 style bumper mount with latches on the drivebase frame that latch onto mounts on the bumper. The second way is a 2017 style bumper mount that uses guiding rods that go into the drivebase frame and some other latching system. The 2017 is definitely a harder bumper mount to work with, however, the straight face with mounting provisions may prove useful if we decide to deploy ramps for our partners to climb up on. We will investigate this in a cartoon CAD.

Task: Drivetrain gearbox design

  • We decided on a triangular configuration of mini-CIMS, with ratios that yield about 10ft/s and 18.6 ft/s. We are now trying to figure out how to package the shifter cylinder into the gearbox below the CIMs. We are exploring either a smaller bore cylinder or a nose-mount piston instead of pancake. Changing from 3/4 to 9/16 bore reduces the shifting force from 22lb to 11lb; Further testing is needed to see if this is enough force to shift. Regular air cylinders can be used if it is ok that they stick out longer than the CIMs. The next steps are to finalize which shifter cylinder we want to use, model the rest of the gearbox plate, and then detail in the entire gearbox assembly with Cheesy Parts.

 

Intake

Task: Test intake prototype with pivotable arm

  • We assembled the intake prototype and mounted it to the 2015 drivebase to test. We were able to run a few tests with a single set of wheels, but unfortunately, as the intake was made out of 0.25" plywood, a part of the arm broke off and we were unable to gather any further data.

Task: Test 2015 Simbotics style intake prototype with pivot wheel sets

  • We initially based the intake prototype off of 2015 Simbotics’; with the first set of wheels pivotable but not the second set–we spaced out the second set to be tangent to the cube when it is 13" width. When intaking a cube in a square position, the intake easily worked, however, when intaking a cube in a diamond position, the cube got trapped between the first set of wheels which were able to pivot to comply to the cube's shape and second set of wheels which were stationary. To fix this, we made the second set also pivotable and were able to intake a cube at different angles. We hooked up the motors on the intake to the drive talons on the development board and were able to simulate different speeds for different sides by turning with the control board joysticks. With different speeds, we were able to force the cube to rotate into a square position in a certain direction which shows that some amount of asymmetry may be necessary to intake cubes. We will continue testing this intake on the field carpet to account for the friction between the cube and field and will further design this in CAD into a stronger prototype.

Programming

Task: RPILIDAR Driver

  • We finished up writing the driver and implementing all necessary methods. The driver is ready to be tested once the LIDAR sensor arrives.  

Task: LIDAR Visualizer

  • We continued work on the LIDAR visualizer. We set up a Node JS server that recieves LIDAR point data encoded in JSON from the roboRIO using Network Tables. The server then uses websockets to send this information over to the viewing page. We also continued work on the visualizer interface. It can now plot fake JSON data, but needs to be hooked up the the websocket so it can display a continuous stream of data from the LIDAR sensor.

FRC Build Blog Day 1 (1/6/2018)

Here is the 2018 game animation.

Game Analysis
This year’s game involved many variables, requiring a strategy that succeeded in performing all tasks presented in the challenge. With the power ups, scoring based on points/sec, randomized sides, and the difficulty of climbing, no straightforward strategy which ensures a win presents itself. After trying various combinations of scores, we concluded the best way to maximize our points would be during autonomous when the scoring is scaled by a factor of 2 and during the endgame when each climb is worth 30 points. Since autonomous scoring is worth more than tele-op scoring, we decided that autonomous scoring was a major priority to stay ahead during a match. We planned for short paths to score cubes (<30 ft), so drivetrain acceleration and our ability to score from either side of the robot was critical for rapid cycles. We debated between starting off on the middle or side of the field. If we were to start off on the middle, we would have an equally distant route to reach the switch and scale. However, we could not guarantee the successful completion of our autonomous route as our alliance partners may get in our way. If we were to start on the side, we may either have a very long or a very short route to reach the switch and scale. However, we could guarantee the successful completion of our autonomous route as our alliance partners would not get in our way. During the endgame, we aimed for having all 3 robots climb or by having 2 robots climb and 1 levitate.

Drivebase
Task: Fix 2015 drivebase and prototype if it can go over the platform ramp

  • We fixed the 2015 drivebase which was missing a control system. We mounted our development board on the drivebase and programmed it to drive, but immediately saw issues as it went up the ramp. Because of the long wheel base and short distance between the wheels and the ground, the drivebase high centered on the ramp when the bellypan came into contact with it. This result shows that for our 2018 drivebase, we will need a shorter wheelbase and a greater centerdrop.

Task: Model drivebase motion up the platform ramp

  • We modelled in CAD the motion of a possible 2018 drivebase going over the platform ramp without high centering. We tweaked two factors (distance of outer wheels from side edge of frame rail, center drop) until we arrived with the dimensions of a drivebase that would safely clear the ramp with 4 wheels (2 on each side) always contacting it and without the bumpers or bellypan ever contacting it.

Intake
Task: Fix 2015 intake and prototype if it can intake Power Cubes

  • We fixed the 2015 intake which was initially missing the intake wheels, shafts, and timing belts. We used this intake to prototype the intaking of the Power Cubes. We did see some promise with this intake, however since it was designed for the tote and not the Power Cube dimensions, we were not able to efficiently intake the Power Cube. Moving forward, we plan on testing out different wheel materials to contact the Power Cube as it is being intaked and will design/prototype an intake with the correct geometry to fit the width of the Power Cube.

Field Prototyping
Task: Test coefficient of friction of Power Cube on scale plate

  • We wanted to calculate the coefficient of kinetic friction between the Textured HDPE and the fabric of the Power Cubes. To do this we looked at how high we had to lift the straight piece of HDPE before the Power Cube started to slide down. From this we calculated the angle formed and the coefficient of kinetic friction which ended up averaging to about 10.32 degrees and 0.18, respectively.

Task: Test geometry of 3 robots on the platform

  • We laid out the spacing of three robots on the scale platform to see if it was physically feasible of having 3 robots side by side with enough space for the robots to maneuver to those locations. It seems that it is possible for 3 robots to fit on the platform.

Task: Find what angle does the scale tilt to at its highest state

  • At its highest state, the scale tilts to around an angle of 7.662 degrees. Based on our data gathered from the task: “Test coefficient of friction of Power Cube on scale plate”, the Power Cubes will not slide down the scale plate when it is tilted to its highest state.

Field Construction
Task: Assemble wooden scale

  • We assembled a wooden version of the scale, however we are planning on assembling an aluminum version similar to the scale we will see in competition. This is because we don’t want any differences from the tournament scale from different materials that will arise in different materials (center of mass, weight, friction) which may impact our robot performance and drive our robot design.

CAD
Task: Space claim of 3 robots on the scale platform

  • In CAD, we made a space claim of 3 robots, placed them in various arrangements on the scale platform, and looked at the feasibility of having all 3 side by side. It seems that it is possible for 3 robots to fit on the platform.

Programming
Task: Program 2015 drivebase to steer in correct direction with controller joysticks

  • When we initially placed the development board control system on the 2015 robot and attempted to drive it, the wheels were turning in the opposite direction to the controller joysticks. We corrected this by reprogramming the drivebase to match with the direction of the controller joysticks.

Game Analysis, Scouting, and Strategy Workshop

On Thursday, December 7th, we hosted a Game Analysis, Scouting, and Strategy Workshop to help acquaint new team members with an important aspect of the robot game. The workshop by led by a member of Team 254’s drive team, Brandon Chuang. During the workshop, we taught new members about the importance of pit scouting and match scouting and how we use the data that we collect in order to form great alliances. In order to give our members a realistic experience, we helped them analyze a 2016 FRC Match Video and track the number of points scored by each team during the autonomous and teleoperated periods. Members were introduced to the team’s scouting database and how it is used at tournaments to organize collected data. Our drive coach, Kevin Sheridan, and FRC Technical Lead, Ashwin Adulla, also helped provide examples of some hardships that Team 254 went through when forming alliances at tournaments. We also emphasized the importance of working together with other teams and how that plays a role in an alliance’s performance. Overall, we strategically prepared our new members for kickoff!

Discovery Day 2017

Event Summary

On November 11th, Team 254 appeared at the Bay Area Science Festival Discovery Day to demonstrate our 2017 season robot Misfire. At the event, we used a few game elements like wiffle balls, gears, and a makeshift wooden hopper to show over 2000 people the capabilities of our robot. In addition, we spoke about our experiences during 2017 FRC season and the challenges we faced. We collaborated with other FRC teams including Team 1868: The Space Cookies and Team 5026: The Iron Panthers to convey the message of FIRST and STEM. This event allowed us to spread our knowledge and experience to our community in the Bay Area.

Team 254 Members showcase Misfire to the audience

Our Experiences

Our experience at Discovery Day not only spread awareness for STEM within our community, but also made an impact on our team. Outreach leader, Yusuf Halabi stated, “My Discovery Day experience this year allowed me to take a new perspective into 254 and our role as an organization. When I first joined robotics as a freshman, I was under the impression that 254 was only good for creating robots. However, my experience at Discovery Day allowed me to realize that serving and supporting the community is just as important as building robots.” Similarly, Team 254 Operator, Themis Hadjiioannou said, “My favorite part of Discovery Day was seeing the joy on children’s’ faces when they saw Misfire shoot into the high boiler. Watching the light in their eyes flash as they observed what a robotics team was capable of struck a chord with me. I thought back to when I was a kid and was inspired by drones and robotics. Discovery Day made me proud to be a part of an organization that strives to bring this inspiration to kids around the world.” Overall, Discovery Day was an outstanding outreach event and we wish to continue making such a profound impact in our community!

Team 254 Driver, Justin Ramirez, helps load up fuel before Misfire demonstrates its shooter

2017 Winners of Chezy Champs!

A Brief Summary

As part of an eventful weekend, we hosted Chezy Champs, our annually hosted offseason FRC tournament at Bellarmine College Preparatory, in our hometown of San Jose, CA. We had a great time jumping back into action with our robot, Misfire, before the start of the 2018 FRC Season. Alongside, FRC Team 1011 Team CRUSH, FRC Team 696, Circuit Breakers, and FRC Team 5104 BreakerBots, we were able to win the tournament!

Team 254 Members pose for a picture with Misfire after winning at Chezy Champs 2017

Event Highlights

General Highlights

Check out this great highlight reel made by RoboSports Network (RSN)!

Exhibition Match

At the tournament, we hosted an exhilarating exhibition match, which included the several teams with highly skilled shooting robots. The standard rule for this match was that the only way to score points was to shoot fuel into each alliance’s respective boiler. The teams which participated in the exhibition match:

The Blue Alliance

  • Team 971 Spartan Robotics
  • Team 973 Greybots
  • Team 1678 Citrus Circuits

The Red Alliance

  • Team 254 The Cheesy Poofs
  • Team 1323 Madtown Robotics
  • Team 3309 Friarbots

Team 1323, Team 3309, and Team 254 work together to gather fuel to shoot into the boiler

The Red Alliance started off with a significant lead by scoring 50 fuel balls into the boiler, making it an amazing show for the audience to see 3 robots scoring fuel into one boiler at once! Eventually, the Blue Alliance slowly started catching up with the Red Alliance by consistently scoring their collected fuel into their boiler. With a minute left in the match, the our friends from Team 973 displayed their true defensive skill by blocking Team 1323’s path across the field. Though the Blue Alliance’s efforts were equal to that of the Red Alliance, the Blue Alliance was unable to keep up with the pace of the Red Alliance. After the great efforts shown by both alliances, the Red Alliance came out to be the winner scoring 203 kPa compared to the 97 kPa scored by the Blue alliance. The Exhibition Match Video can be viewed here.

A Special Thanks

Chezy Champs was very special this year due to the efforts of so many people and organizations. Team 254 would like to thank our friends from RoboSports Network (RSN) for providing our audience with such a great analysis of each match and team at the tournament. We would also like to thank all the volunteers, who helped make Chezy Champs possible. It was truly an incredible experience to host an offseason tournament attended by so many talented teams, even those who chose to attend from out of state, and to make new friends and catch up with old ones too!

Robot Performance

Throughout the qualification matches, we encountered a few mechanical problems with our gear grabber, hopper, and drivetrain and a few problems with Misfire’s autonomous performance. After some careful observation, we realized that our gear grabber’s knife-edge was worn out, so we replaced it with a new piece of polycarb. Our stationary hopper panel kept colliding with the field hoppers, which tore our hopper wall. We replaced the broken hopper panel to solve this problem. We also noticed a crimp lodged between our chain and sprocket on our drivetrain, so we removed it to make our drivetrain function normally. Overtime, our autonomous performance improved after our mechanical changes. Thus, we were able to seed first for alliance selection after having a qualification match record of 7 wins and 3 losses.

Misfire successfully hangs within the last few seconds of a Qualification Match

During elims, we faced some fierce competition during our Semifinal and Final Matches. We ended up having to compete in a third tiebreaker match in semifinals and finals. In Semifinals Match 2, the opposing alliance of Team 973 Greybots, Team 1538 The Holy Cows, Team 604 Quixilver, and Team 2135 Presentation Invasion, scored a large amount of kPa during auton and kept their lead up by activating all 4 rotors! In Finals Match 2, the opposing alliance of Team 1323 Madtown Robotics, Team 3309 Friarbots, Team 5026 Iron Panthers, and Team 2073 EagleForce, displayed their true skill, by keeping a consistent lead, from the start of the match. By the end of our elimination matches, we won all of our tie breaking matches, allowing us to win the event with our alliance. We wouldn’t have been able to accomplish such a victory if it weren’t for our amazing alliance partners – Team 1011 CRUSH, Team 696 Circuit Breakers, and Team 5104 BreakerBots.

Members of the Team 254 Pit Crew repair Misfire before an upcoming Finals Match

Fall Project 2017

Team 254 is kicking off the school year with a Fall Project. This year students will be building a gear robot for the 2017 FIRST Steamworks challenge. The goal of this year’s Fall Project is to provide students with the opportunity to pursue whichever aspect of FRC they want to in a stress free environment, in order to prepare them for build season in January, and to have fun! Team 254’s Fall Project robot will be competing in Madtown Throwdown this coming November.

2017 Technical Resources Release

Team 254 is proud to present the Technical Binder, Code Release, and Blog Release for our 2017 World Champion winning robot, Misfire. If you have any questions, feel free to comment on the Chief Delphi thread for our code release and our technical binder release.

Festival of Champions

2017 FRC Class of FIRST Champions

Recently, we attended the first ever FIRST Festival of Champions in Manchester, NH. After winning the St. Louis Championship with our extraordinary alliance partners, we faced the Houston Championship winning alliance in an intense series of 5 matches. Alongside Team 2767 Stryke Force, Team 862 Lightning Robotics, and Team 1676 The Pascack PI-oneers, we were able to win the Festival of Champions Event. We enjoyed a tour of DEKA and a great potluck at Dean Kamen’s home. We would like to say thank you to FIRST for giving us the amazing opportunity to participate in this event!

“Thanks to the phenomenal performance by our alliance partners!”

Championship Matches

In the most intense games we’ve played all season, the Houston alliance won the first two matches. Our alliance won the third and fourth to stay in the game. In Match 4, our alliance partners helped us win against the Houston alliance by a margin of only 16 points. In the final match, our alliance capped a great season with one last win. We won the match, and the festival championship, setting a record high score for this year’s game of 588. We’ll let the video speak for itself, as our last finals match was certainly one of Team 254’s best performance of the season.

We would like to honor our alliance partners because of how far we all came, from St. Louis to Festival of Champions. We would also like to thank Team 973 Greybots, Team 5499 The Bay Orangutans, Team 1011 CRUSH, and Team 2928 Viking Robotics from the Houston Alliance for setting the bar high and making all our matches an intense, and great experience. We’re excited to meet other FRC teams at our annual offseason event, Chezy Champs, and to kick off next year’s season!

2017 World Champions!

This past week, we attended the FIRST Robotics Competition Championship in St. Louis. With two Championship events this year, it was hard to decide which to compete at, but we had a great time facing off against teams from the east coast, Canada, and more. Alongside, Team 2767 Stryke Force, Team 862 Lightning Robotics, and Team 1676 The Pascack PI-oneers, we were able to win the Daly Subdivision finals before continuing on to win the Championship at Einstein.

Daly

In our qualification matches, we went 7-3 overall and seeded second, with Team 2767 holding first seed. The improvements we made since the Silicon Valley Regional payed off as we hit over 100 kPa during our second qualification game and managed an average of about 381 points per match. Throughout qualifications, we were also regularly able to achieve 40 kPa and activate 4 rotors, giving us enough ranking points to propel us into a high seed.

Moving into alliance selection, we joined Team 2767 alongside Team 862 and Team 1676. With a balance of gear and fuel scoring ability, we planned to maximize our points by hopefully achieving 40+ kPa and activating 4 rotors every match. Throughout the Daly finals, we remained undefeated and won against incredibly fierce competition. In our first semifinals match, we managed to score a personal record of 550 points! We were also very grateful to have received the Innovation in Control Award, an award we had also received back at the San Francisco Regional. After winning Daly with some very close matches, we were anxious to see how we would perform at Einstein.

Einstein

In a round robin style tournament, we competed against every other subdivision champion throughout 5 matches. We ended up going 3-2 here and seeded second, giving us the opportunity to compete in the grand finals as the blue alliance against the Darwin champions, who had beaten us in the round robin tournament.

In the most intense games we’ve played all season, our alliance pulled ahead in the finals, winning the second match by only 3 points. We’ll let the video speak for itself, as the second finals match was certainly many on 254’s favorite game of the season:

We are so grateful to our alliance partners for playing such a huge role in our success, and to every team that attended the Championship for competing with us and making it such an amazing experience. Special mention to the Einstein teams for going all the way and challenging us in ways we hadn’t previously been. We’re looking forward to facing off against the Houston alliance at the Festival of Champions!

“Thanks to our mentors for such a great season!”

“Post-Einstein excitement”

FRC Day 42 Build Blog

Day 42: Preparing for World Championships

Quick Summary

Everything that we need at competition, which includes the pit, was packed and prepared for being shipped to St. Louis.

Robot Progress

Shooter

We worked on some hood geometry and started to look at the placement of the hanger tube and different sizes. We also finished the design of the new hood and CNCed the new hood plates. We laser-cut a new center divider for the hood. Everything was assembled at on Monday’s build.

Hopper/Intake Ramp

The old hopper broke on the practice bot, so we laser cut new plates and assembled them onto the robot. We also cut spare hopper plates and spare intake ramps in case we need them during champs.

Driver Practice

We started practicing a new kind of gear cycle called the check mark cycle, which we plan will probably use from here on out.

Image uploaded from iOS (3)-min.jpg

Programming

Today, we worked on shooter tuning and semi-automated gear placement in teleop using an ultrasonic sensor.  For shooter tuning, we worked on fixing a problem where our stream will begin to walk backwards over long periods of time.  We believe this is a result of the system becoming more efficient over time as the bearings on our flywheel heat up.  To solve this problem, we tried compensating for the increased flywheel speed over time by taking an average speed of the flywheel between shots and using it to adjust our voltage..  For teleop gear placement, we tried to use the ultrasonic sensor to limit the amount the robot can drive backwards when scoring gears to prevent it from driving too far back and bending the spring.  However, we found that the ultrasonic sensor updates far too slowly for this to actually work right now, so we are investigating ways to speed up the ultrasonic sensor and use our drivebase encoders to compensate for this lag.

FRC Day 41 Build Blog

Day 41: Improving Gear Scoring


Gear Grabber

Today we worked on the Cow Catcher, or an assembly that prevents balls from getting stuck in the Gear Grabber and pushes balls out of the way. We began with prototyping with two delrin side plates and a 3/8" standoff, and while the results were promising, balls still got caught in the general vicinity, affecting its performance. To solve this issue, we prototyped the cow catcher with a longer standoff, which prevented balls from getting caught.

Inline image 1

Having determined the relative effectiveness of the Gear Grabber, we began to design a deployment mechanism such that the cow catcher fits in the framer perimeter at the start of the match. We came up with the use of a torsion spring and hardstop to deploy the cow catcher at the start of the match. Due to the limited availability of springs at the lab, we used rubber bands instead, but the deployable cow catcher was ultimately too weak. Thus, we added standoffs and reinforcing plates to increase rigidity.

Inline image 2

Robot Progress

We worked on fixing up our practice bot, making a new intake ramp and straightening out our hopper flanges. We also did a lot of shooter testing and were able to significantly improve the accuracy of our shots.

Silicon Valley Regional Tournament

This past weekend, we competed at the Silicon Valley Regional at San Jose State University.

Qualifications

Throughout the qualification matches, Misfire performed very well, despite bearing one loss, going 8-1 overall and scoring an average of about 300 points. We focused on a strategy of reaching 40kPa along with delivering a couple gears to maximize the amount of ranking points we would receive, which worked out as we were seeded first heading into eliminations



Eliminations & Awards

In our alliance, we picked Team 604 “Quixilver Robotics” and Team 4990 “Gryphon Robotics.” With this alliance, we aimed to focus both on fuel and gears, hoping to reach 40kPa+ in the boiler, 4 rotors spinning, and all robots hanging in an ideal game. During quarterfinals, we beat the prior record of 506 points with no penalties by scoring 507 points and later in semifinals we scored 509 points. During finals we finished off with scoring 522 points (517 without penalties)!



Through the eliminations, we ended up winning the tournament after an exciting final match! We won the Quality Award and were recognized for being captain of the winning alliance. It was a great experience playing against all these teams and we hope to see some of them again at World Championships in St. Louis.



FRC Day 40 Build Blog

Day 40: Improvements for SVR



Driving Improvements

We added additions to our robot in order to create an easier and more efficient driving environment. We have placed a blue LED strip on the back of the hood to let drivers know whether or not they are in a good shooting proximity. To do so, we wired up MOSFETS to control the LED strip, which will plug into the DIO port on the roborio.

Inline image 3


CamerasfAutono

At SFR, we found that installing “back-up cameras” would help with gear placement. Therefore, we worked on optimizing the camera placement, as well as finding a way to widen the narrow field of view (currently 70 degrees) to better capture the gear’s location.

Inline image 2

Autonomous

After the post-mortem we worked to improve the robot's turning during autonomous. We prototyped small Delrin sliders that didn't quite work when sliding on carpet. We anticipate that the best solution will be to try to implement drop-down omni wheels that we can turn on. However, if that does not work either we will resort to changing the autonomous routine itself.

Inline image 1


San Francisco Regional Tournament

This past weekend, we competed at the San Francisco Regional at St. Ignatius College Preparatory.

Qualifications

Throughout the qualification matches, Misfire performed strongly, going 10-0 overall and scoring an average of about 277 points. We focused on a strategy of reaching 40kpa to maximize the amount of ranking points we would receive, which worked out as we were seeded first heading into eliminations.



Eliminations & Awards

In our alliance, we picked Team 971 “Spartan Robotics” and Team 4990 “Gryphon Robotics.” With this alliance, we aimed to focus heavily on the high goals, hoping to get 40kpa+ in the boiler, 2 rotors spinning, and all robots hanging in an ideal game. We also planned to use defense, having one of our robots block the chokepoint near the gear loading station, hopefully making it impossible to for other teams to score 4 gears.



Throughout several exciting rounds, we ended up winning winning the tournament after an intense final match. We also won the Innovation in Control Award and Griffin Soule, our team president, won a Dean’s List Finalist Award. It was a great experience playing against all these teams and we hope to see some of them again at SVR.



Misfire Reveal

Team 254 presents our 2017 entry into the FIRST Robotics Competition: Misfire. Misfire will be competing at the inaugural San Francisco Regional, followed by the Silicon Valley Regional and the St. Louis FIRST Championship. More information on the robot.

FRC Day 37-39 Build Blog: Preparing for Competition

Days 37, 38, & 39: Preparing For Competition


Intake

Today, we worked on fixing the intake after it broke earlier this week. After analyzing the strength of our current weldment and modifying it slightly, we decided that simply strengthening it would not help the fundamental problems existing in the design. We took one of the broken intakes and attached it to new polycarb side plates hoping that this would decrease impact forces by increasing the time of impact (change in momentum = force * time). We drove around and crashed the intake several times with minimal damage occurring, and we are proceeding with the new design.

Inline image 1

Hopper

Today, we worked on the Hopper and testing the fully assembled redesign. After fixing some initial discrepancies in the real life assembly, we tested out deployment and stowing. As of a few tests, the intake deployed well (after modifying the intake ramp slightly to have slightly deeper notches to fit the constraint screws in) and we will be updating the CAD with these changes on Friday. We will also laser cut a new intake ramp to get definitive results. If all systems work well after that, we will begin assembling the new parts for comp as well as a back up.

Between, Friday and Saturday, we were able to finalize the hopper design. After some CAD work on the redesign and some troubleshooting after assembly, we were able to make a solid hopper. More testing will continue to confirm consistency, but we will be moving forward with the Comp set soon.

Shooter Flywheel

In the last week, we explored the use of a large flywheel to help increase the consistency of the shooter. A flywheel dramatically increases the kinetic energy of the shooter wheel system, hopefully reducing the amount that the wheel slows between shots. We designed a steel flywheel with about 4 lb*in2 of inertia. It is geared up 2.5:1 from the shooter wheel to increase the kinetic energy. Preliminary testing shows that adding the flywheel may not actually achieve the desired result. The elasticity of the timing belt caused some controllability challenges that the programming team worked on this weekend. Each time the direction of tension in the belt switched (from energy going into the flywheel to energy leaving the flywheel), there was a little bit of elastic movement which added oscillations to the shooter wheel system. The plots looked like the typical mass/spring/damper plots for a 2nd order system from a controls or dynamics class.

Climber

We are using inline elastic to add stretch to the rope. This will allow the robot to pull the rope down with a low amount of force at first, getting some wrap before beginning to lift the robot off the ground. This is important because the Velcro that engages the climber drum is strong in shear – and its strength is proportional to surface area. Once the rope has wrapped around the drum enough, the load transfer switches to Capstan forces. We are assembling several ropes, each with a number of pockets like the ones shown below.

A video is linked below of the climber being tested. Note that in this test case, we were pulsing the motors. Typically, the climb is much faster.

https://youtu.be/UYMntdI0CPo

Gear Grabber

The latest iteration of the gear grabber has a dual pivot design – the whole mechanism pivots from a pickup configuration to a scoring configuration with one pivot, and a second lower pivot is sprung to allow the wedge to conform to the floor without digging in. We built this prototype last week out of delrin on the laser cutter and used it for driver practice over the weekend. We moved forward with making this out of metal and sent the parts to Anodize on Monday morning.

Check out the video below for pickup and scoring performance of the current prototype:

https://youtu.be/Oj6FkjkXHtM


Driver Practice

We practiced driving by cycling gears with and without defense, climbing, and intaking balls off the floor. We are much better now at quickly scoring gears, and in one practice match we scored 9 gears and climbed when no defense was played on us. The gear grabber is much more reliable, and consistently picks up gears off the ground the second we touch them. Scoring is still tricky, but we are getting better at finding the right depth at which we need to drive into the spring.

Programming

Autonomous Program

Today, we worked on finding and fixing a bug in our Adaptive Pure Pursuit controller.  The bug was causing arc lengths in the controller to be calculated incorrectly in some cases, resulting in a value much higher than it should have been.  This would in turn cause the robot to drive way too fast and shoot past the end point in the path.

Gear Grabber State Machine

Today, we worked on implementing a new Gear Grabber state machine. The new gear grabber code greatly simplifies the operator controls. Now, there are only 2 buttons used for intake and release. The state machine uses output current to determine when a gear has been picked up and will activate the piston automatically.

More Information

We also found two bugs resulting in the robot not turning left and twisting along an autonomous routine, and we were able to fix these edge cases.

The hardware team began crimping and attaching 15 battery leads and running battery tests to determine true battery capacity.

FRC Day 36 Build Blog

Day 36: Improving Subsystems


Hopper Floor

Today we ordered all the the parts for the upper extension of the hopper floor and got majority of the CAD done. We will be going with 50 degree edges made out of Delrin with the first two rollers of the upper extension not having any surgical tubing to allow for horizontal movement.

Feeder

Today we worked on inventorying all the COTS parts for the Feeder subsystem. This would allow us to become more organized and make sure parts we need we have or are buying. I also helped lead students in inventorying other subsystems via the CAD,

Inline image 3

Gear Grabber

We worked on the gear grabber. We removed the polyurethane added before with black, conforming WCP wheels. We found them to work significantly better than the polyurethane, but still not extremely well. We continued to work and changed the WCP wheels to red conforming wheels, which while ugly, worked better.

Inline image 1