Meet Our 2024 RoboSubs

With the next generation of Team Inspiration now partaking in RoboSub, our goal was to leverage past experience and mentorship to accelerate the learning curve of new team members. Learning by doing, our team members further developed Onyx and Græy and aim to utilize inter-sub communication to help optimize our subs’ performance on the competition course.

Season Recognition

5th Place Overall Standing
3rd Place Team website
6th Place Technical Paper
8th Place Team Video

Specs

Technical Design Report

Onyx Pre-qualification Video

Graey Pre-qualification Video

With a new team of six middle and high school students ranging from Sarasota, FL to San Diego CA, we had less experience, knowledge, and time than most other competing teams. We focused mainly on quickly learning our existing capabilities and optimizing Græy and Onyx.

Learning from last year’s results, we focused on improving our perception and localization abilities, utilizing a sensor fusion system including hydrophones, a Fiber Optic Gyroscope (FOG), Doppler Velocity Logs (DVLs), and Inertial Measurement Units (IMUs), combining these with a camera-based perception pipeline. We continued to improve our mission planner, incorporating situational-based input rather than relying on sequential commands. We decided to deploy our more capable AUV, Onyx while reserving Græy as our backup AUV that would be deployed for our stretch goal of demonstrating basic intersub communication.

We intend to deploy Onyx to complete all six missions & intersub communication, while Græy functions as our test bed for developing code. Onyx, equipped with a marker dropper, torpedo launcher, gripper, and hydrophones, is capable of completing all missions.

Competition Strategy

Græy has acrylic enclosures for the battery and the electronics . Græy, like Onyx, uses an 80/20 frame and has a double O-ring seal around each enclosure. Græy’s thruster configuration is identical to Onyx’s. While Græy is not equipped with servos, its similarities to Onyx make it an ideal testbed for software development. Since intended to be deployed to demonstrate intersub communication, Græy also has both a soft and a hard kill switch.

Mechanical

Onyx is our main sub and has two acrylic battery enclosures and one electronics enclosure. Each enclosure contains a double O-ring system to ensure waterproofing. The frame consists of 80/20 extruded aluminum beams, a design that was chosen due to its low cost, simplicity, modularity, and scalability. 80/20’s lack of hydrodynamics is not a concern in RoboSub, in which AUVs move at low speeds. Containing eight thrusters arrayed for holonomic drive, a marker dropper, gripper, and torpedo launcher, Onyx is equipped to attempt every mission. To meet safety standards, we have two kill switches: a soft-kill to cut power to the thrusters, and a hard-kill to cut power to the main computer.

Marker Dropper: The team is reusing its marker dropper from RoboSub 2023. Through analysis and a trade study comparing different variations of marker dropper, the final dropper was previously decided to be modeled after a design from KyuTech’s AUV. The mechanism is 3D printed and uses a Blue Blue Trail Engineering SER-2000 Underwater Servo Enclosure with a goBILDA 2000 Series servo inside.

Gripper: Our gripper uses a Newton-Subsea-ROV based actuator with modified jaws. The jaws were designed so that the gripper can reliably hold mission elements of various shapes. Mounted on a magnetic linear slide rail, the gripper is easily deployed on the bottom of the sub before each run.

Torpedo Launcher: We designed a torpedo launcher utilizing a spring as the main propulsion element. The spring is 3D printed and its latch mechanism is powered by an underwater servo. This mechanism sends a torpedo that is reliable enough to reach the intended target.

Custom Enclosure

Both Onyx and Græy have custom acrylic enclosures for both battery and electronics housing. Onyx has a +50% larger electronics enclosure volume than Græy - Onyx’s enclosure being 18.25 in. long with an 8 in. diameter, while Græy’s enclosure is 11.75 in. long with an 8 in. diameter.Onyx has a battery enclosure on both the port and starboard sides, while Græy has a single battery enclosure under the electronics housing.

Both AUVs utilize Jetsons as onboard computers. Onyx uses a more powerful Jetson Xavier NX, while Græy uses a Jetson Nano. To enable autonomous movement, both subs use a Pixhawk 4 as flight controllers. Onyx, with its larger electronics enclosure, is expandable. Onyx contains a custom hot-swappable Power Distribution Board (PDB) which enables us to connect two batteries to the PDB at the same time. This allows us to replace a low-voltage battery with a new one without shutting down the system, giving us more than 1.5 hours of continuous runtime. Onyx also has a custom Printed Circuit Board (PCB) for processing hydrophone signals. On both AUVs, the external sensors communicate with systems in the main electronics enclosure via connectors which are sealed from the inside and the outside of the main end cap, rather than simple penetrators that can fail with weak epoxy.

Electrical

Overview

Electrical Configurations

Onyx’s Electrical Diagram

Græy’s Electrical Diagram

We equipped both of our AUVs with forward-facing and downward facing cameras. On Onyx, we have a front-facing USB low-light, front-facing OAK-D Wide, and bottom-facing OAK-D Wide, while Græy is equipped with a forward-facing OAK-D Wide and a bottom-facing USB low-light. OAK-D cameras divert processing power from our onboard computers, allowing us to easily run YOLOv8 models for object detection. OAK-D Wides have the two additional benefits of having an increased field of vision and an onboard IMU unit, which we can use as sensors to determine relative heading. We switch to USB low-light cameras when using OpenCV color thresholding and other OpenCV methods as they allow for higher resolution images in low-light conditions.

Cameras

On both subs, we have USB low-light, OAK-D, and OAK-D wide cameras. The OAK-D Wide cameras have Inertial Measurement Units (IMU), as do the Pixhawk controllers. Onyx has a Fiber-Optic Gyroscope (FOG), Teledyne DVL (Doppler Velocity Log) and hydrophones, while Græy has a Water-Linked DVL.

Sensors

Hydrophones

In our custom hydrophone board we include 3 daughter boards, which connect to 3 Aquarian Scientific AS-1 hydrophones. Each daughter board includes a pre-amp, a custom amp, a custom variable frequency filter, adjustable between 10khz to 40khz, a digital potentiometer for variable gain, and a custom active high-pass filter to ensure the signal is as clean as possible. A Teensy 4.1 board takes in the signals from the hydrophones to save processing power. The signals are analyzed using algorithms that include time of arrival and signal phase to triangulate a specific heading for an acoustic pinger. This heading is then given to the main processor on the AUV.

We utilize the Robotic Operating System (ROS), compatible with C++ and Python, to facilitate communication between components. Our computer communicates with our flight controller, the PX4 (Pixhawk) via MAVROS, a special kind of ROS node that enables communication between MAVLink flight controllers. We utilize sensor feedback loops, through perception/cameras or external sensors to send motion values and servo values to our actuators.

Software

Mission planning is broken into three parts – mission decision, navigation to the mission, and a perception feedback loop upon mission localization. Our high level mission planner uses logic gates and history of success to determine the mission sequence. Once a mission is determined to be executed, lower-level scripts are used to navigate to the approximate area of the mission, after which a perception loop utilizes cameras to identify objects and move the AUV accordingly to complete the mission.

Mission Planning

To track translational motion, Doppler Velocity Logs (or DVLs) are mounted onto both subs. DVLs contain four cells that emit sound waves, and the change in frequency of the echoes, which return from the ground, are used to calculate the velocity of the sub. Velocity is integrated over time to calculate position. This information is used to aid station keeping and to navigate to a specific position (ex: the buoy location). The Fiber Optic Gyro (or FOG) is used to obtain information about the sub’s heading. The FOG is more accurate than other onboard compasses because it is immune to electromagnetic interference.

Navigation

For efficient and accurate object detection programs, we use YOLO based data training. During the training process, we utilize augmentation filters, hundreds of images, and roughly 400 epochs. This ensures the model can identify the desired object under variable conditions and environments. To ensure accurate object detection, models are tested in a Jupyter Notebook via Google Colab. This allows us to run the model against sample images and predict the certainty of object detection. Once the models are tested, we integrate them into our software. We also utilized color detection as a second layer of identification. If visibility is too low for the YOLO model to be effective, the color detection algorithm assists with computer vision dependent tasks.

Computer Vision

Testing Strategy & Version Control

We strove to ensure individual components worked before integrating them into our AUVs. We tested each script and sensor individually when possible to ensure reliability before integration. If it was not possible to test sensors/code without being mounted/connected to our AUVs, we bench tested to decrease risk and save time. Unit testing allowed us to minimize unknown points of possible failure and maximize efficiency, which was critical with our limited time.

We utilize GitHub as our storage and development platform for software. Widely considered the most convenient cloud-based software development platform in the world, GitHub provides a platform for convenient version control and code sharing that can track changes made by team members. With constant changes to our code by multiple members, and dozens of different scripts in multiple languages, it is important to quickly identify which team member changed which script and which lines, in case the latest version(s) are nonfunctional. GitHub offers all of these abilities, making it the ideal version control system for software development.

System Specs

Weight

With 1 battery: 55.6 lb

Degrees of Freedom

3 translational (x,y,z)

3 rotational (roll, pitch, yaw)

Power System

Lithium-ion battery 14.8V, 15.6Ah (x2)

Wireless Communications

Succorfish Delphis modem

Wifi

Dimensions

24.5”w x 31”l x 19.5”h

Endurance

Normal usage: 1.5 hour runtime

Main Processor

Jetson Xavier NX

Propulsion System

T200 thruster (x8)

Full throttle FWD/REV thrust @ nominal (16V) 11.6 / 9.0 lb f

Sensors

Blue Robotics low-light camera

OAK-D Wide (x2)

Fibre Optic Gyro (FOG)

Imaging scanning sonar

Navigation System

Pixhawk orange cube

Teledyne Xplorer DVL

Software Architecture

Robot Operating System (ROS)

Payload

Gripper

Torpedo launcher

Marker dropper

Acoustics

Hydrophone array of AS-1 Hydrophones (x3) and custom digitally adjustable notch filter circuit