Our Projects

Over 10 major projects across 3 unique subteams

Airframe

Camera Gimbal

Camera Gimbal

Airframe

The airframe team collaborates with the embedded team to work on the camera gimbal. The 2-axis camera gimbal is used to keep the camera in the plane stabilized and facing downwards towards the ground at all times, even as the plane performs pitch and roll maneuvers in the air. Currently, the camera gimbal has 3D printed PLA gimbal arms wrapped in carbon fiber to increase the stiffness and ensure that in case a piece breaks, a replacement part can be quickly 3D printed.

Fiber One

Fiber One

Airframe

Fiber One is a custom built, all composite plane with a 10.5 ft wingspan. Its namesake comes from the fact that it is built from fiberglass and carbon fiber. The plane was designed, built, and tested by the airframe team. The Fiber One has a tapered, swept, high wing design and a conventional tail. However, the Fiber One is unique in the fact that it doesn't use landing gear which adds weight and drag. Instead, the Fiber One takes-off from a detachable cart and belly lands with protective skids on its belly.

Manufacturing

Manufacturing

Airframe

The airframe team uses a combination of composite manufacturing, 3D printing, and laser cutting to produce components for the plane. Composite manufacturing is used for the wings, fuselage, and tail because composite materials have a high strength to weight ratio. Smaller components or components with complex geometry are manufactured using 3D printing for easy and fast reproducibility. For example, the camera gimbal arms and unmanned ground vehicle chassis are 3D printed. Laser cutting is used when manufacturing planar components such as bulkheads or core material for spars.

New Plane Design

New Plane Design

Airframe

The team is currently in the process of discussing the configuration of a new flight platform. The team has developed a MATLAB program that takes in a desired wing span, cruise speed, and weight and calculates an ideal wing geometry by optimizing the plane’s efficiency. Utilizing analysis software such as XFoil and Ansys, the team will be able to select the most efficient designs for the airfoil and tail. The airframe team will also work with both embedded and software teams to ensure the design will be capable of performing all competition tasks with ease.

Swallow

Swallow

Airframe

The Swallow is our current main flight platform purchased for competition. The plane was modified by our team in order to perform all tasks for the competition. It features a mid-wing design and a V tail with a modified tail wheel for steering during take-offs and landings. It also has a custom power system and a reinforced landing gear area in the fuselage to accommodate the increased weight from the onboard electronics.

UGV

UGV

Airframe

The unmanned ground vehicle is one of the newer and more challenging additions to the competition. It must drop from the plane, land safely on a targeted area, and drive to another location, all autonomously. Minimizing weight and volume are critical for conserving valuable payload space, and optimizing the plane’s flight performance. Currently, the team has a prototype for the chassis, and is working on implementing an autonomous system to guide the unmanned ground vehicle to the target while it is in the air.

Embedded

Antenna Tracker

Antenna Tracker

Embedded

The antenna tracker holds the WiFi antenna at the ground control station that communicates with the plane. The team designs and implements the firmware for this in-house built antenna tracker, which enables the team to employ its members more efficiently during missions. This firmware compares the antenna tracker’s location and the plane’s location to compensate for a direct physical path between the plane and antenna leading to better transmission across these devices.

Autopilot

Autopilot

Embedded

Currently, the embedded team works with a Pixhawk 2 Cube Black as the flight controller for autonomous flight with the plane, which communicates with the ground control station over RF to send telemetry data about the plane. The team sets up the firmware for the flight controller and ensures that it interfaces well with the control surfaces, the motor, and communication systems. In addition, the team works on tuning the autonomous mode of the flight controller through several configurations of PID values.

Onboard Systems

Onboard Systems

Embedded

Amongst the plane’s payload there is a full size DSLR camera, a compact computer, a camera gimbal, and a long range WiFi antenna. The team mainly uses these devices to fulfill both the autonomous and object detection task of the competition. For the most part, the team uses the WiFi network for transmitting large packages between the plane and ground control station, such as the pictures which the camera takes and sends back, and the path the team’s software generates. The compact computer serves as a hub for distributing these packages to the appropriate device.

Power and Signals Board

Power and Signals Board

Embedded

Both the power board and signals board are in-house designed and built circuit boards tailored to efficiently run the power and communication systems of the plane. The power board uses switch-mode power management integrated circuits to power the different devices inside the plane at regulated voltages. The signals board uses various digital devices, such as microcontrollers, multiplexers, and analog-to-digital converters to manage most of the signals running across the plane.

Software

Computer Vision System

Computer Vision System

Software

  • Tools : Python, PyTorch, Jupyter
  • Skills : Deep Learning, Computer Vision

While traversing the search area, the system must automatically detect and classify ground targets. To accomplish this task we develop a robust four part computer vision system: saliency, segmentation, classification, and filtering. In order to accomplish this task with both speed and accuracy, the team utilizes various neural network frameworks such as PyTorch, TensorFlow, and YOLOv4. We are currently in the midst of streamlining our entire computer vision pipeline by creating a central computer vision server that will manage each subtask and communicate between our other systems.

Ground Control Station Backend

Ground Control Station Backend

Software

  • Tools : Golang, InfluxDB, Grafana, Docker
  • Skills : Backend Web Development

The backend web server of the ground control station communicates between the other systems of the software team. It is also responsible for maintains connections with the AUVSI SUAS Interoperability System during competitions. In addition, the backend collects telemetry data from the plane where it is displayed on easy-to-read graphs using Grafana.

Ground Control Station Frontend

Ground Control Station Frontend

Software

  • Tools : ReactJS, Javascript
  • Skills : Frontend Web Development, UI/UX Design

The ground control station allows for real time monitoring of the mission and its associated tasks. Both the computer vision and path planning systems are integrated into the GCS to create a centralized hub that can directly interface with each component asynchronously. The goal is to provide an interface that completely abstracts the implementation of the backend so that anybody can manage the mission directly via the website.

Path Planning

Path Planning

Software

  • Tools : Python

During the competition, our plane must accurately hit a series of waypoints and cover a specified search area while avoiding stationary and dynamically moving obstacles in the field. To accomplish this task we develop a robust path planning system that uses Dubins curves to plan obstacle free paths, and also modify the waypoint navigation controller of the Ardupilot autopilot to perform dynamic obstacle avoidance. This year’s competition has updated the dynamic obstacle to be another team’s UAV that could be flying concurrently with ours. We are planning on improving our system to include live search area coverage and adversarial UAV awareness.