Projects

See what our teams have been up to!

Learn more

Artificial Intelligence

Building off of a Robot Operating System architecture, AI tasks range from integrating COTS peripherals, to object detection, navigation, top-level task reasoning, and simulating the boat’s performance in a virtual competition course.
Tech stack: Docker-based environment, Git for version-control, ROS Kinetic for the architecture, and C++/Python for the majority of the source code.

CV/LIDAR deep learning

The goal for CV deep learning is to implement deep learning models to accurately assign various classifications to objects detected by our boat’s ZED camera. These models will replace our current classical CV detection system, which is slow and unstable. Upon completion, this new detection method will be integrated into our boat’s architecture, along with the development of a streamlined process for creating and training new models. Members will gain hands-on experience with one or more steps of the deep learning process, including model selection, design, training, and validation, as well as data annotation and preprocessing. Additionally, members may optionally also seek to gain a more theoretical understanding of deep learning models (and by extension, machine learning as a whole).

LIDAR quantification

The Lidar quantification project’s goal is to explore metrics to evaluate our current LIDAR detection algorithm. These metrics give a more precise, fine-grained measure on the quality of the LIDAR detection than our current method of using a visualizer to holistically observe the results of the detection. Members will gain experience with researching and conceptualizing various evaluation metrics, then evaluating the “informativeness” of each metric using a small, annotated 3D point cloud dataset. Potential metrics include those typically used in the fields of machine learning or data science, as well as custom metrics that we may formulate ourselves.

Task Planning

The task planning team acts as the brains of our boat - it analyzes perception data in order to direct the movements of path-planning and controls. This will include creating an advanced movement library that interfaces with the path-planner/controller, enabling our boat to strafe, activate anchor mode, maneuver the boat relative to an obstacle, and circle around an obstacle. Expanding the boat’s current high-level logic capabilities, this new library interfaces with perception, path-planning, controls, and simulation. Task-planning members will achieve a broad understanding of all of the boat’s sub-teams. This team is great for someone who likes abstract, critical thinking without having to spend lots of team reading through ROS documentation, academic articles, or complicated math.

Task Sim team

The Task Sim team’s goal is to build a simulated version of the Roboboat competition tasks with Gazebo in order to test our boat code without an actual boat or water. This will involve testing the boat code as a single unit in its ability to traverse course obstacles, as well as constructing metrics to quantify and compare the performance of specific sections of the boat. Since we have a working simulator and some foundational code for path verification from previous semesters, existing code will be used and improved upon. Members of this team will gain experience with each section of the codebase, from perception to motion-modelling, by interfacing with and testing them, as well as experiencing simulation/physics engines.

Electrical

The Electrical Team is responsible for the allocation of power to the boat and management of cables and wires from the battery to the computer, sensors, and motors.

Electrical Box Reorganization

Our main focus this year was to make the box more organized, neat, and reliable. To achieve these goals, we introduced numerous new parts and custom PCB’s. For example, this year we purchased a breakout board, which splits an ATX cable into its derivative voltage supplies. We also overhauled nearly every single electrical splice connection and adopted a more crimp-oriented approach as opposed to soldering everything.

Hydrophone Array and Processing

We use an array of 4 hydrophones in a rectangular fashion to find the pinger for acoustic docking. Our design for the hydrophones is to use a Raspberry Pi as our processing power and then communicate via USB to the main computer. In order to reduce interference, the hydrophones are placed at the very bottom of the boat. They are placed in a rectangular fashion to determine both the direction and the angle of the pinger, as compared to a straight line in which the pinger would have 2 possible locations.

New Electrical Box

Over the past couple of years, we have been utilizing custom-cut polycarbonate to construct the main “shell” of the electrical box. This year, we used a pelican box - a rigid, waterproof, and air-tight crate. In order to drill holes in the box for some inputs and outputs, we needed to expose members to different tools needed to penetrate the thick material. Although this process was time consuming, all of the sub-team members were able to contribute, learn a new skill, and the electrical box was completed almost two months ahead of where it was last year. We also built our electrical box based on a CAD model designed and completed last summer.

Hulls & Systems

H&S Team works on the design and construction of the team’s boat. This includes designing around the boat’s electrical system, accompanying drone, and sensor specifications required by the boat’s autonomous systems.

Trimaran Hull Design

This year, we made some innovative changes to our original monohull design. We changed our hull design to a trimaran design to increase boat stability. Specifically, it reduced downward pitch when going forward, which was caused by the additional weight due to sensors and mounts at the front of our boat.

Trimaran

Vacuum Infusion

In previous years, the team utilized pre-resin-impregnated, high-temperature curing carbon fiber cured in an autoclave to construct the hulls. This year, the hulls were constructed using a carbon fiber vacuum resin infusion process. Each demi-hull was fabricated using a machine-milled high-density foam mold produced for the team by the Ford Motor Company.

Vacuum

Auxiliary Vehicle Team

The Auxiliary Vehicle Team is responsible for the development of an Unmanned Aerial Vehicle (UAV). Our UAV has typically been a drone. Tasks include the Computer-Aided Design (CAD) for the physical structure for the drone, designing the electrical and sensor system that the drone uses, and developing the communication interface with the boat.
Tech stack: ArduPilot framework, Python MAVLink API for controls.

Drone Design

A custom designed H-frame was fabricated to increase stability in flight and allow for a space-efficient payload delivery mechanism. Due to COVID-19 restrictions, reduced working capabilities prevented aluminum fabrication as was seen in previous iterations of the team’s UAV solution. Instead, prototyping and final fabrication of all components excluding 4 carbon fiber frame tubes were manufactured using the team’s 3D printer.

Business

The Business Team secures funding for the team, processes reimbursement requests, maintains sponsor relationships and is responsible for social media engagement. The subteam also creates and maintains the team website.

Reimbursement Request Tracking

Previously, our reimbursement process involved uploading receipts to a shared drive and notifying an Authorized Signer in the team. When preparing the team budget, we realized how challenging and tedious it was to match the receipts with SOAS account records of reimbursements. To solve this problem, we created a Google Form to help with the recording and tracking of reimbursement requests.

Members: Ryan Do

Reimbursement

Social Engagement

Last year, the Business Team was able to organize more social engagement events with the funding opportunities provided by the Central Student Government (CSG) for student organizations. We organized a virtual Burger Night in Fall 2020 and reimbursed members who participated for the cooking ingredients. We also purchased Minecraft licenses for team use and a Realms subscription, pathing the way to Minecraft game nights in Winter 2021.

Members: Sim Yi Lin, Dora Guo, Rimaz Khan, Ryan Do, Ryan Draves

Social

UMA Website

This year, we decided to focus on improving our previous year's website by updating the site with up-to-date information, and creating new pages (Projects and Blog pages) and sections to provide more information about the team. This current website is created using Jekyll. We have also created Figma design layouts and started some work in creating a new website using GatsbyJS. We chose static site generators with the intention of Search Engine Optimization (SEO). The new GatsbyJS website would also provide more control compared to the templated Jekyll site we currently have.

Members: Sim Yi Lin, Dora Guo, Rimaz Khan

UMA