The 29th annual Intelligent Ground Vehicle Competition (IGVC) was held this past June in Rochester, Michigan. This competition will be a familiar classic to our alumni; RoboJackets has been in attendance every year since 2003, with the only breaks being in 2005 and 2020 due to a lack of manpower and the COVID-19 pandemic, respectively. However, this year’s IGVC marked a bittersweet end to this succession, as the RoboNav team had decided that this would be their last time in attendance. Instead, efforts will pivot towards the University Rover Competition (URC). With that in mind, the team went into this year’s IGVC ready to make some memories and put Swervi to the test. 

IGVC consists of two distinct parts: the autonav challenge and the design competition. The autonav challenge is the robotics test of IGVC, with teams’ autonomous vehicles competing to complete (or get as far as possible through) a course consisting of barrels, painted white lines, and ramps or other miscellaneous obstacles. The design competition gauges team-specific innovations and improvements through a presentation and question-and-answer session with the judges. 

Caption: Swervi ready for battle at IGVC 2022

For this year’s competition, the new swerve drive robot Swervi was finally up to bat. The ideation and mechanical design of Swervi started in 2019-2020 when Jessiii was completed; however, the robot had an extended development time due to the cancellation of the 2020 IGVC. Aptly-named Swervi’s swerve drive design was a step outside the box for RoboNav, since nearly all previous robots built for IGVC had differential drivetrains. It also marked a turning point in the team, as RoboNav had found success with Jessi and its iterations over the past four years. Stepping into brand-new territory presented a great learning opportunity but also resulted in brand-new challenges.

In the lead up to IGVC, the electrical team was still working on firmware for swerve drive motor control. In contrast to Jessiii, which only had two DC motors to control via one motor controller, Swervi had four DC and four stepper motors which required more controllers and different types of controllers. Due to this added complexity, Swervi arrived at competition essentially paralyzed. The team was working on debugging and improving the control during the competition by using stepper motors with individual controllers; however, the mbed microcontrollers were eventually burnt out and broken. This resulted in the team failing to qualify for the autonav challenge, which was a big disappointment after working  so hard on Swervi over the past year. Charles Li, the 2021-2022 project manager of RoboNav, said he felt it was a letdown for past members, especially during the team’s final IGVC. “I thought about the other components of Swervi, how long ago [the] design started, and how many people worked on it.…I couldn’t help but feel a bit disappointed that their efforts couldn’t amount to what we wanted it to amount to.”

Caption: The team works on Swervi at competition

However, all was not lost. The team still presented the completed Swervi design to the judges and placed 3rd in the design competition. Additionally, the team bonded and grew closer by facing the ups and downs of competition over 4 days. Priyanka Rajan, the upcoming project manager, stated she was surprised by how much she was able to get to know her teammates over the course of IGVC. Despite the disappointment in the autonav challenge this year, the future looks bright for RoboNav.

The team leads present the Swervi design to the judges

RoboNav receives the design award for placing 3rd.

RoboNav bond at the apple cider mill after a hard day at work

Speaking of the future, next year RoboNav will take one small step for RoboJackets, one giant leap for robot-kind. The team will fully focus their energy on URC after spending the last two years researching and prototyping in preparation for the new competition. There is still much to do, including the establishment of and recruitment for a new Science subteam to tackle the astrobiology aspect of the competition. Nostalgic alumni may be disappointed that the team is setting aside IGVC after all this time, but do not fret. Electrical lead Andrew Roach said it well: “We’re moving on from IGVC but we’re not forgetting about our past experiences… We’re trying to build on and learn from [what we worked on in the past few years for IGVC].” 

So, after bonding and commiserating over all the good and the bad of IGVC 2022, RoboNav is moving together towards a brand new challenge to continue learning about and advancing in the field of robotics.

Phillip recently moved from El Segundo, California, to Seattle, Washington, where he is currently helping open a second office as part of his role of site director at Second Order Effects, an electrical engineering consulting firm. Outside of work he enjoys hiking and backpacking, and is looking forward to exploring the PNW area this summer. He also took up beer brewing during the pandemic.

We asked him about the lessons learned for RoboJackets that continued to serve him after college. He responded that he had learned so much, “from how to manage budgets and basic accounting, to conflict resolution and leadership.”  The skills that he felt had helped him the most were “effective communication and collaboration with a team… [which included] everything from how to delegate responsibility to presenting information to leaders in the institute. When I started working professionally, I felt I had a leg up in communication skills.”

Finally, we asked Phillip if he had any favorite memories from his time at RoboJackets. He highlighted RoboCup 2008 in Suzhou, China, saying: “It was our first international trip and while we didn’t do great in competition, it was a lot of fun to plan and participate in. We even got funding from a local company to pay for our tickets!”

The goal of the competition is to have go-kart scale autonomous head-to-head racing. As the competition is still relatively new and building up to head-to-head racing, the current format is the winner is determined by the fastest time completing 5 laps, increased from 3 laps last year.

One of the biggest visual changes to the robot was the addition of “The Box”, originally intended to be a large plastic box that contained most of the electronics and cabling. The single container design allowed for the consolidation of at least 3 enclosures and approximately 15lbs in weight saving, being incredibly helpful with general movement of the robot and controls. Additionally, it greatly reduced wiring complexity and allowed the robot to be much easier to waterproof, with racing in light rain a very real possibility at this competition. Unfortunately, due to supply chain related delays, the box had to be made out of wood. This did eliminate the easier waterproofing, but retained other benefits of the centralized electronics enclosure.

 At EVGP 2021, the motors used for steering and braking were stepper motors that burnt out due to underestimating the torque required to steer/brake. To correct this problem on the new revision, a mechanism redesign to reduce torque paired with a switch to brushless motors for more power (and torque) helped both the steering and braking theoretically be controlled on the robot. New custom PCBs interfaced with the ODrives, allowing a transition to STM32 for more advanced controls and interfaces. The entire robot was also rewired to take advantage of the single consolidated electronics plate, easing maintenance and improving grounding/signal integrity, from the box addition.  

 When the team started the week-long competition on Tuesday, some issues were identified with the ODrives, the motor controllers used for controlling both the steering and braking systems. The ODrives, while incredibly customizable, are difficult to debug and can be easily damaged. Debugging and replacing the ODrives consumed many competition days (and nights) of the members in attendance. Once the steering and braking systems were electrically working around Friday, it was unfortunately still understeering compared to what software was commanding, making it very difficult to complete the road course track with both direction turns.

The software team has been transitioning to ROS 2, which includes a brand new high level control planner. This re-write is still in progress, so the existing planner was used at competition. Using the existing planner, paired with the understeering issues, caused some technical performance issues with the robot such as always going in reverse. Additionally, some parts of competition were spent configuring the wireless communications system to work with the track network, an important precursor to multi-robot racing.

On race day on Friday, there were still some steering and braking issues. This led to the robot only going forwards and backwards, resulting in the fourth place finish. 

Sam Walters, the former RoboRacing PM, said “​​Our main takeaway is that we were too ambitious with our goals and didn’t put in the hours appropriate to reach them in time.” The team will be spending the time until next competition doing minor improvements to aspects that were close to working, redesign aspects that didn’t make the performance cut, and trying to get the robot in a testable state for software much sooner in the year. Doing a complete redesign allowed for the team to do a lot of learning in such a short period of time, effectively having two competition seasons in a single year! The team is excited to put this vast knowledge gained to work on the next version of the robot!

Some alumni may remember that RoboJackets has brought combat robots and RoboRacing robots, including Bigoli, out to Colorado to compete at Sparkfun AVC (Autonomous Vehicle Competition). Sparkfun has since stopped hosting that competition, but this past May a new AVC took its place: the Angry Vehicle Competition. This new AVC only has combat robotics in 6 weight classes: plastic ant (1 lb, with limited materials), antweight (1 lb), beetle (3lb), hobbyweight (12 lb), featherweight (30 lb), and sportsman (30 lb, with kinetic energy limits). This year, RoboJackets competed at AVC with three featherweight bots: Anarchi, Chonkii, and Binari.

Anarchi is a horizontal spinner with an asymmetric bar weapon. Anarchi had 2 losses and one win. Anarchi’s first fight was against Wedge Ham, a wedge bot that was able to control the match effectively and went on to win the event in first place. Anarchi lost by tap-out after it was unable to move. Its second fight was against Dragonfruit, a Huge-style vertical bar spinner. Anarchi won this fight after it knocked one of Dragonfruit’s large wheels clean off, prompting a tap-out. Anarchi’s final fight was against Crippling Depression. Unfortunately, by this point in the competition, Anarchi was havins issues with its drive systems and lost the fight by knockout after getting slammed into and stuck in the corner of the arena.

We reached out to Brian Naing, one of the builders of Anarchi and the current BattleBots PM for commentary on Anarchi. Brian was pleasantly surprised at Anarchi’s single victory, because the team was down to the wire on completing the bot:”The first real battle was finishing the bot in the first place as we only got our bot’s weapon up and spinning an hour before we had to leave the shop for the airport.”

Overall, Brian was pleased with the durability of the chassis, but saw many areas for improvement in the robot’s serviceability. Of the chassis, Brian said “[it] was a brick which held together well enough to take the constant ramming from a wedge bot (that later went on to win AVC) while being low to the ground enough to be virtually invincible to an over-cutter bot we went up against.” The weapon “spun-up decently well at first but would have probably been more powerful with some modifications.”

The major lesson that the team learned was to focus more on the details: “the main things we learned at AVC were that small design details can literally make or break the bot and that it’s more important to make sure the basic subsystems of the bot are designed well to be reliable.” Additionally, Brian noted that repairability was a major area for improvement: “we had not designed the bot with the intention of these components being replaced when damaged which led to a design that required half of the bot to be disassembled in order to just access and repair the gearboxes.”

Next up is Binari, a shuffling undercutter with two independently driven weapons, enabling the robot to undercut no matter its orientation. Binari unfortunately had a 0-2 win/lose ratio. Its first fight was against Crippling Depression, another undercutter. It delivered some solid hits to Crippling Depression, but Crippling Depression was able to land a hit on the bearing system on one of Binari’s weapons, disabling one of the weapons for the rest of the competition.

Binari’s second fight was against HUHMWPE DUHMWPE, a horizontal spinner made primarily of UHMWPE. HUHMWPE DUHMWPE was able to tear several pieces of  Binari off, including essential shuffler plates, ultimately resulting in a KO.

We reached out to Knute Broady, Binari’s bot lead for comments. Although Binari saw stiff competition, Knute said, “we did have a bot that worked as we intended when we entered into the arena, and in many ways that is a success.” The big thing that Knute was proud of in the design was the shuffling system:

The part of the design that was the most work, the shufflers, actually ended up working pretty darn well. The main thing we struggled with at the comp was an ESC issue that should have simply been avoided, but in the end they were maneuverable and robust. We suffered a ton of damage in our first match, but despite bent plates and shafts we were able to get the shufflers to work again which says a lot about them as a design. In fact, the robot could still move even after the side plate had been completely ripped off.

According to Knute, the major issues with the robot’s design was in the weapon systems: “The Achilles heel of the Binari design actually turned out to be the weapon belts.” Binari used polyurethane weapon belts that were shredded by internal features (including the weapon motor that had ribs) and caused multiple weapon failures. Additionally, “our ‘overcutter’ [IE, the second weapon that was mounted above the robot] was too high and didn’t allow us to hit either of the bots we were facing … Our bot could actually accommodate a double undercutter, however we were never able to try it in competition because of sheared bolts after our first match.”

Similar to Anarchi’s team, Knute’s big takeaway was to refocus on the basics:

Overall, Binari had a lot of good ideas, but it’s hard to execute such an ambitious bot when we missed on a lot of the basics. I think the failure of Binari also reinforced focusing on an engineering first approach rather than a ‘it should work’ approach that it’s easy to slip into in battlebots. It wasn’t that fact that we had 2 weapons or shufflers that caused us not to be successful, it’s the little things we missed along the way.

Last up is Chonkii, a shell spinner shuffler. Chonkii is back after its last appearance at Motorama, this time with a 1-2 win/lose ratio. Chonkii’s first fight was against Booster, a flipper. It was able to deliver a handful of hard hits on Booster, including one that bent its flipping forks, resulting in a win by tap-out. Chonkii’s second fight was against Wedge Ham. Wedge ham was able to push Chonkii around, ultimately pushing Chonkii up onto the side of the arena. It got stuck there and lost by KO, despite its best efforts to spin itself free. Chonkii’s last fight was against Yahoo, a rematch after Yahoo flipped it over and KO’d it at Motorama. Unfortunately, Chonkii ran into drive issues in the fight, and lost by KO, once again, flipped up on its back.

We reached out to Mihir Nagaraj, Chonkii bot lead for comment. The major change the team made since Motorama was changing how their shell was constrained: “We fixed the bearings and shaft constraints  that broke at motorama causing our weapon to stop spinning. Our rematch against Yahoo showed that these changes (changing the bearing material to steel lathing a slot into the center shaft) fixed this issue.”

Mihir attributes the electronics failures to poor electronics securing: “Our weapon system worked well but we learned that our electronics need to be better secured as our drive system lost power and lost our last fight.” The major lesson learned was that the team needs to pick higher quality materials (E.G. steel for bearings) to be more competitive.

Overall, the team had some good fights and learned some good lessons at AVC, summed up by one of Brian’s comments: 

The main things we learned at AVC were that small design details can literally make or break the bot and that it’s more important to make sure the basic subsystems of the bot are designed well to be reliable first before trying to implement a novel, game-changing design or pumping as much energy as possible into a giant weapon. I think Anarchi and other teams as well have also taken home the idea of designing for reliability and repairability like many other teams at AVC did. Any bearing will always have the chance of breaking upon an impact meaning that the only countermeasures are trying to spec the bearing right, making it easier to pull out if you have to, and bringing a bunch of spares. I think this, along with alot of other small and big things Robojackets has learned from this competition will be immensely valuable as we build new and better bots next year.