Boat Overview (Software)
New Upgrades On Minion For The 2024 Competition
GB-CACHE flowchart with required processes in blue and optional processes in yellow
GB-Cache
GB-CACHE can efficiently segment and extract concave hulls from unstructured point clouds by sub-sampling these point clouds to a structured grid. In addition to segmentation and hull extraction, the technique is designed for the optional integration of point filtering, object classification, mapping, and object tracking processes.
Yolo v8
Minion’s newly optimized video pipeline allow users to deploy trained network models and leverage deep learning for a variety of tasks. The modular software approach gives team members more chances to participate in the process of creating custom neural networks, like a YOLOv8 model trained to detect light towers!
Controls System
The software controlling Minion's propulsion system enables precise autonomous navigation through an all-new control strategy. Using a feedback-driven control system, it dynamically adjusts thrust and steering for optimal movement across surge, sway, and yaw. High-level navigation commands are translated into specific motor actions via an inverse kinematic model, enabling agile path following and station-keeping. The software features multiple operational modes, automatically selecting the most efficient setup based on current conditions. Safety protocols ensure continuous operation in case of faults, providing reliable and autonomous functionality across diverse environments.
ROI Extractor
The precise calibration between Minion’s cameras and LiDAR sensors allows the LiDAR-based object detection to extract a region of interest (ROI) from the camera video stream. A 3D volume is ‘drawn’ around the object and projected onto the image frame to define the ROI in 2D. Custom software enhancements ensure that camera and video data streams are synchronized which enables real-time color detection and image-based object detection with YOLO.
