Autonomous surface vehicles sound conceptually simple: a boat that drives itself. The reality is a deeply integrated system where mechanical design, embedded electronics, sensor physics, and software algorithms must all cooperate perfectly — in an environment that is wet, salty, and unpredictable. This article is an engineer’s tour of the Minion platform as it existed heading into competition, subsystem by subsystem.
For a higher-level overview of how the platform fits into our competition history, visit our boat overview.
The Hull: WAM-V Foundation
All RobotX teams receive a Wave Adaptive Modular Vessel (WAM-V) hull from competition organizers. The WAM-V 16 is a catamaran-style inflatable pontoon craft roughly 5 meters long with articulating sponson attachments that allow the hull to flex with wave action rather than pitching rigidly through it.
The WAM-V’s flat equipment deck is both a blessing and a constraint. It provides ample mounting area for electronics enclosures and sensor masts, but the open structure means every connection, enclosure, and cable run is exposed to spray, UV, and salt. Weatherproofing becomes an engineering discipline in itself.
Structural Modifications
Team Minion added a custom aluminum rail system along the centerline of the equipment deck. This modular mounting rail — machined in-house at ERAU — allows sensors and enclosures to be repositioned between runs without drilling new holes or fabricating new brackets. The time saved during competition preparation is significant when a task briefing might require rearranging sensor geometry.
Propulsion
The Minion platform uses two electric thruster pods mounted at the stern sponsons, driven independently. Differential thrust — varying the output of each thruster — provides steering. There is no mechanical rudder.
Why Electric?
Combustion outboards are lighter for equivalent power, but they introduce vibration that corrupts IMU measurements, exhaust that foul air quality sensors, and complexity (fuel, mixing ratios, cold-start reliability) that competes with the team’s attention during competition. Electric drives are quieter, cleaner, smoother, and the battery state is always known precisely — a meaningful advantage when station-keeping tasks require sustained thruster activity.
Motor controllers communicate over a CAN bus to the main autonomy computer, receiving torque commands at 50 Hz. This update rate is sufficient for the vessel’s low-speed maneuvering envelope while leaving processing headroom on the bus for other devices.
Sensing Architecture
Minion’s sensor suite is designed around the principle of complementary modalities: each sensor class covers the failure modes of the others.
LIDAR
A rotating 2D LIDAR mounted on the forward mast provides 270 degrees of obstacle awareness at ranges up to 40 meters. The sensor’s scan plane is tilted slightly downward to intersect the water surface ahead of the vessel, catching low-profile obstacles that a higher-mounted scanner would miss.
LIDAR returns are processed in real time into an occupancy grid that feeds the path planner. Dynamic obstacles — objects that move between consecutive scans — are flagged and tracked with a lightweight Kalman filter. The LIDAR sensor array has been one of the most iterated components across platform generations.
Stereo Camera
A calibrated stereo camera pair handles buoy detection, color identification, and light pattern recognition. The stereo baseline gives depth estimates out to roughly 15 meters — useful for approach path planning when the task requires precise standoff distance from a target.
Image processing runs on a dedicated GPU-equipped compute module rather than the main mission computer, keeping perception latency low regardless of what the planner and controller are doing. Computer vision and object detection capabilities have been central to competition scoring in every RobotX event.
GPS and IMU
Position is derived from a dual-antenna GPS receiver that provides not just position but also heading — independent of the magnetometer, which is unreliable near the thruster motors. The IMU provides high-rate attitude and acceleration data for the controller and for filtering GPS position during momentary satellite dropouts.
GPS and IMU data are fused using an Extended Kalman Filter (EKF). The EKF implementation is a standard component of ROS’s robot_localization package, validated against known-good trajectories in our test pond before each competition season.
Hydrophone Array
For acoustic beacon tasks, Minion carries a custom four-element hydrophone array suspended beneath the hull. Time-difference-of-arrival processing across the array elements provides a bearing and rough range estimate to an underwater acoustic pinger. This subsystem is detailed in depth in a separate technical article.
Power Systems
Battery Configuration
Power is supplied by a lithium iron phosphate (LiFePO4) battery bank chosen for its thermal stability and tolerance of partial state-of-charge cycling. Unlike lithium cobalt oxide chemistries, LiFePO4 does not enter thermal runaway easily — a critical safety property for a vessel operating in a public waterway at a competition event.
The bank is sized to support full competition runs plus a reserve margin. A battery management system (BMS) monitors cell voltages and temperatures, providing an emergency shutdown if any cell exceeds safe limits. Power system design and marine battery selection are covered in detail in complementary technical documentation.
Power Distribution
A custom power distribution board routes 24V from the battery bank through individual fused branches to each subsystem. The kill-switch required by RobotX rules interrupts all thruster power while keeping the autonomy computer and communications live — allowing the team to observe and diagnose the vessel’s state after an emergency stop without losing telemetry.
The Autonomy Stack
The software architecture follows a layered model:
- Hardware abstraction — ROS drivers for each sensor and actuator, providing a uniform interface to higher layers.
- Perception — Object detection, localization, mapping.
- Planning — Global path planning (waypoint sequencing) and local path planning (obstacle avoidance).
- Control — Trajectory tracking and station-keeping controllers.
- Task executive — Behavior tree coordinating which tasks to attempt and in what order.
The behavior tree executive was the most important architectural decision in the 2016 rebuild. It allows tasks to be added, removed, or reprioritized without touching core control code — a property that becomes invaluable when competition briefings reveal task changes 24 hours before the event.
Lessons That Shaped the Design
Every subsystem described above exists in its current form because something failed in a previous iteration. The LIDAR tilt angle was chosen after an untilted installation missed a partially submerged log during a test run. The LiFePO4 chemistry was selected after a lithium polymer pack swelled in Florida summer heat. The modular rail system exists because a sensor bracket had to be relocated between tasks using only the tools on the dock.
This is what engineering in a competitive environment actually looks like: continuous, specific, failure-driven improvement. The Team Minion community has carried that culture through every generation of students who have contributed to the platform.