Meet Kirby
Mechanical
Fig. 1 Expanded Sub View
Main Structure
The mechanical structures on Charybdis consist of the hull, propulsion systems, and additional sub-systems (such as the torpedo launcher). The hull design can be divided into two main sub-components:
1.) Outer Frame: Provides overall structure and stability and the
2.) Inner Frame: Comprising the structures that hold the housing tube, battery tubes, and computer components.
The hull is constructed from 6061-T6 aluminum plates. The plates were cut on a water jet and slotted together.
Inner Housing
All of the subs control electronics are housed in a large 8-inch acrylic tube to centrally house the sub’s critical controls. The sub also has four smaller acrylic tubes mounted to the end of the outer frame. Three of the tubes hold the subs 1000mAh batteries with a fourth tube for auxiliary system circuitry.
For propulsion, Charybdis is powered by eight T-200 Blue Robotic Thrusters. The thrusters are laid out in the Blue Robotics vector 6DOF configuration which gives the sub maximum maneuverability.
Fig. 2 Main Tube
Fig. 3 Battery Tubes
Torpedo System
One of the sub-systems we integrated with Charybdis is our torpedo launcher. The torpedoes are launched using a compressed CO2 canister. The launching mechanism uses a waterproof servo with a gearbox that translates the rotational motion to linear motion. A 3:1 ratio is used to amplify the linear force on the torpedo firing pin which triggers the canister release.
Fig. 4 Torpedo Flow Simulation
Fig. 5 Torpedo Launcher
Electrical
Main Structure
The electrical team focuses on creating an electrical system that had not only new features on the sub but also to make or redesign parts of the system to help with their performance and quality of life. Different software programs like Eagle CAD, LTspice, and Altium help develop and test out the designs of new circuits for the electrical system. Along with that, members are taught how to design circuit boards and solder components to help grow the team and the community.
Power Board
The sub utilizes a power distribution board that is divided into multiple parts. The first section connects the killswitches to 5 lithium polymer batteries which then connect to the motors, onboard computer, and sensors. The second section is managed through the kill switches that were designed for sending power to the rest of the electrical sub systems.
Fig. 6 Power Board
Kill Switch
The use of kill switches is to manage the power distribution of the sub by acting as a way of killing all power to the sub via a switch on the back of the sub.
The current kill switches utilize recalculated tracc widths to prevent overheating and current overload. In addition, the current kill switches were designed for the physical geometry of the board to be directly attached to the chassis of the chamber along with bringing all of the existing components closer together. This change eliminated excess height and width that was previously seen and allows for more electrical components to be on the frame of the sub.
Fig. 7 Kill Switch
Software
Software Architecture
The software architecture of Charybdis is built on a foundational structure utilizing the Robot Operating System (ROS), which provides a message-passing system, networking capabilities, and more. The packages we created were designed to take advantage of ROS and the open-source libraries that use it, including SMACH, MavROS, ROS Serial, OpenCV, and Tensorflow. Figure XXX shows the high-level overview of Charybdis and signals direction.
Fig. 8 Software Architect
State Machine
High-level decisions about what Charybdis should do are made in a state machine implemented in SMACH, a ROS package that defines a state machine structure. Each state performs a specific task and is strung together to make more complex processes. Video input is received from two USB cameras, one forward-facing, and one downward-facing, and sent to the object detection algorithm via ROS. Once the network visualizes the detections, Charybdis can perform movements based on that information. The Neural Network takes two points from the field of view: one provided by the SSD (Single Shot Detector) architecture within TensorFlow and another provided by the center of the camera. The program calculates the error between the two points and processes the error through a PID control loop, then outputs an RC value published to MavROS. We use this information to calculate sub-movements and feed into our state machines.
Simulations
In addition to the issues we face with sub functionality, the COVID-19 caused increased difficulty with team communication and system testing. We decided that in order to mitigate these obstacles testing of Charybdis’ states and functions prior to an in-person test would be crucial. Thus we decided to use Software-in-the-Loop with Ardusub, using MavROS to send communications. This allows us to get a visual 2D simulation running of Charybdis and to test functions and states before executing them on the physical sub in water. This has streamlined the software design process and improved the speed of implementation for new behaviors.