Every competition season starts with a decision: patch or rebuild. After reviewing what went wrong in Singapore, Team Minion chose a third path — targeted, disciplined improvements to the subsystems that most limited performance, leaving proven components untouched. Here’s what changed for 2016, and why.
Why We Didn’t Start From Scratch
The temptation after a difficult debut is to throw out the old design and start over. It’s emotionally satisfying and sometimes genuinely necessary. But it’s also expensive in time and institutional knowledge. The Minion hull, power distribution architecture, and basic sensor complement had all performed within acceptable parameters in Singapore — the failure modes were more specific than that.
We identified three primary gaps:
- Sensor mounting inflexibility — repositioning a sensor between tasks required tools and time we didn’t have
- Controller tuning gaps — station-keeping failed under environmental disturbances the controller hadn’t been tuned for
- Software architecture brittleness — adding a new behavior or fixing a bug in one module risked breaking another
The 2016 upgrade program addressed all three without touching the hull or power systems.
The Modular Rail System
The most visible hardware change for 2016 was the introduction of a continuous aluminum T-slot rail running the length of the equipment deck centerline. Individual sensor brackets and enclosure mounts attach to this rail via captured T-nuts that can be repositioned with a single wrench.
Why This Matters in Competition
RobotX competition weeks involve rapid configuration changes. Task geometry can change between briefings and the actual event. A sensor that was optimally placed for one task may create self-occlusion problems for another. Before the rail system, moving a sensor mount meant drilling into existing aluminum extrusions and hoping the new holes didn’t weaken the structure.
With the rail, the team can reconfigure the sensor layout in under ten minutes. During the 2016 Hawaii competition, this capability was used twice — once to move the forward camera further from the LIDAR after discovering occlusion in pre-competition testing, and once to rebalance the deck weight distribution after adding a piece of equipment we hadn’t planned for.
The rail design was drawn up by a mechanical engineering student who had zero prior boat experience. That’s a feature, not a bug — fresh eyes on a problem often produce the most elegant solutions, unburdened by assumptions built up over years.
Material Selection
Marine aluminum alloy (6061-T6) was chosen for the rail and associated brackets. It is corrosion-resistant in salt water, machinable with standard shop equipment at ERAU, and light enough that the total deck hardware weight increase versus the 2014 configuration was under two kilograms. All fasteners were replaced with 316 stainless steel to eliminate galvanic corrosion at aluminum-to-stainless interfaces.
Sensor Changes
Forward Camera Upgrade
The 2014 stereo camera pair was replaced with a hardware-synchronized unit offering a wider field of view and lower rolling-shutter distortion at the vessel’s operating speeds. Rolling shutter — the smear artifact that appears when a sensor with a sequential-scan architecture images a moving scene — was causing subtle distortions in the buoy detection pipeline that were degrading range estimates.
Hardware synchronization ensures both camera frames are exposed at exactly the same instant, which is essential for valid stereo depth computation. This single change improved median buoy detection range by approximately 30 percent in test pond evaluations.
LIDAR Mounting Angle
The 2D LIDAR was remounted with a 12-degree downward tilt versus the near-horizontal 2014 configuration. This repositioned the scan plane to intersect the water surface at a distance of 8–15 meters ahead of the vessel — the most useful obstacle detection range for the vessel’s typical operating speeds.
At low tilt the LIDAR’s scan plane passed over many potential obstacles at water level. The angled mounting came at the cost of some long-range coverage, which was an acceptable tradeoff given that RobotX courses are compact and obstacles are never more than 20 meters from the task boundaries.
Controller Improvements
The station-keeping controller received the most intensive redesign effort of any software component.
Integral Windup Limits
The original PID-based station-keeping controller had no anti-windup mechanism on the integral term. Under sustained wind disturbance — exactly what happened during the Singapore docking task — the integrator accumulated a large error value that caused overshoot and oscillation when the disturbance finally relented. Integral windup is a classic control systems problem covered thoroughly in control theory literature.
Adding a simple clamping limit on the integral term eliminated the overshoot behavior in simulation. On-water tests at ERAU’s waterfront confirmed the fix translated to the real platform.
Wind Feedforward
A lightweight wind estimation module was added that uses the difference between GPS velocity and IMU-integrated velocity to infer wind direction and approximate speed during steady-state runs. This estimate is fed forward to the thrust allocation as a disturbance rejection term, reducing the load on the PID’s integral channel.
The feedforward contribution is blended conservatively — it reduces disturbance, but the PID remains the primary controller. If the wind estimate is wrong, the PID catches the error. Defense-in-depth for controllers, the same principle as redundant sensors.
Integration Testing
The upgraded platform went through three phases of integration testing before shipping to Hawaii:
Phase 1 — Component bench tests: Each new subsystem was tested independently before installation. Camera synchronization was verified with high-speed targets. Controller modifications were validated in simulation.
Phase 2 — On-water regression tests: The full system was run on ERAU’s local waterway to verify that none of the changes had introduced regressions in previously working behaviors. Navigation channel transit and obstacle avoidance were specifically re-validated.
Phase 3 — Environment stress tests: The team induced wind disturbances using a fan array and verified station-keeping performance under controlled conditions. Edge cases (full wind + cross-current) were tested explicitly.
This methodology — borrowed from formal systems engineering practice — is described as part of ERAU’s engineering curriculum and is a discipline that Team Minion actively passes on to new members.
What Stayed the Same
Equally important: what the team deliberately chose not to change. The WAM-V hull, thruster pods, battery bank, power distribution board, and GPS/IMU hardware all carried over from 2014 unchanged. This wasn’t laziness — it was a recognition that these systems had not been failure causes and that redesigning them would introduce new unknowns without addressing known problems.
The philosophy of minimum necessary change is one of the harder disciplines to maintain when a team is enthusiastic and creative. It’s also one of the most important. To see the full Minion platform as it came together across all generations, our boat overview traces every design branch.