Boat Design and Technical Information
At the 2025 RoboBoat Competition, The Orca makes its debut as UM::Autonomy’s redesigned autonomous vessel, built from the lessons of two seasons with The Phoenix. Centered around an all-new X-Bow monohull, Orca delivers improved seakeeping and maneuverability to meet the competition’s demands for precise navigation and reliable docking.
To match this new platform, the team consolidated and standardized its electrical systems for resilience and faster iteration, introducing deeper monitoring over CAN and simplifying power distribution and diagnostics. On the software side, we transitioned to ROS 2 and added wind compensation backed by a matured simulation workflow and rigorous weekly in-water testing, creating a robust, modular foundation for repeatable success across core tasks.
The vessel utilizes a 5-foot carbon fiber monohull with a 15-inch beam, scaled up from the "Orca" platform. A new longitudinal fin keel maximizes roll stability in wave environments, while an internal polycarbonate chassis provides a rigid mounting surface that lowers the center of gravity for optimal hydrodynamics.
Visual data is processed by an upgraded YOLOv11 neural network. To ensure robustness, the team leverages Masked Diffusion Models to generate photorealistic synthetic training data, allowing the autonomy system to learn from diverse, computer-generated scenarios that are difficult to replicate in physical testing.
The perception stack features a targetless LiDAR-camera calibration system. Instead of manual markers, the software uses AI feature detection to align natural scene geometry between the LiDAR point cloud and camera feed, ensuring precise sensor fusion and depth estimation.
Dual antennas provide sub-degree heading accuracy independent of magnetic sensors. This mitigates errors caused by environmental magnetic anomalies, ensuring reliable orientation for navigation tasks.
Two Blue Robotics T500 thrusters are mounted vertically in a differential thrust configuration. Positioned deeper and further aft than previous designs, this setup maximizes turning leverage (moment arm) and eliminates ventilation (air intake) for higher propulsive efficiency.
A high-performance Ubiquiti Rocket system ensures reliable, 500Mbps low-latency communication between the vessel and shore control.
The compact enclosure prioritizes thermal management and modularity. By transitioning to integrated PCBs, the design minimizes internal wiring volume, improving airflow and allowing for rapid component swapping during maintenance.
A central 22.2V, 222Wh lithium battery serves as the primary power source, positioned low in the hull to maintain proper vessel trim and stability.
Custom boards replace bulky wiring harnesses for motor control and power distribution. These integrated circuits reduce physical footprint and failure points, while serving as a hardware-level emergency stop to cut power if the remote signal is lost.
The VectorNav VN-300 provides the precise state estimation required for the new "Memory" architecture, which persists object locations in a global map to aid relocation. This data feeds a Dynamic A planner* (smac_planner), allowing the vessel to generate continuous, kinematically feasible paths that account for the boat's turning radius.
Featuring an Intel i5-12450H processor and 16GB of RAM, this compact computer delivers superior performance with a highly efficient 5-Watt idle draw. Its robust integrated graphics accelerate the real-time inference required for the upgraded YOLOv11 and Masked Diffusion pipelines, while a 1 Terabyte drive stores extensive competition logs.
To ensure reliability, a Docker-based environment configured with ROS 2 guarantees identical software behavior across simulation and hardware. This setup supports the team's C++ and Python codebase, managed via Git, allowing for seamless integration across different operating systems.