Undergraduate Projects
Engineering and CS projects for college students. Expect to work with real industry tools, write substantial code, and produce results you can show in interviews.
32 projects
Combine with other filters →Domains at this level
CFD Analysis of an Airfoil with OpenFOAM
Simulate fluid flow over a wing and validate against published data
Set up and run a complete CFD simulation of airflow over an airfoil using OpenFOAM. Generate pressure distributions, velocity fields, and validate your results against wind tunnel data.
Satellite Orbit Propagator
Predict where any satellite will be, minute by minute
Build an orbit propagator from scratch that takes TLE data and predicts satellite positions. Implement Keplerian mechanics and J2 perturbation in Python or MATLAB.
Aircraft Detection from Drone Imagery with YOLO
Train a real-time object detector to spot aircraft from above
Train a YOLO object detection model to identify aircraft in aerial and satellite imagery. Build a complete computer vision pipeline from data labeling to real-time inference.
Parametric UAV Design in Fusion 360 + OpenVSP
Design a drone from requirements to 3D model with aero analysis
Use OpenVSP for parametric aircraft geometry and aerodynamic analysis, then bring the design into Fusion 360 for detailed CAD. Learn the professional aircraft design workflow.
Remaining Useful Life Prediction with scikit-learn
Predict when a turbofan engine will fail before it does.
Build an end-to-end machine learning pipeline on the NASA C-MAPSS dataset to estimate turbofan engine remaining useful life (RUL). You will engineer features from raw sensor streams, train regression models, and evaluate prognostic accuracy using RMSE and scoring functions from the prognostics literature.
Wing Planform Optimization in XFLR5
Squeeze every count of drag out of a wing using panel methods.
Systematically explore the design space of wing planform parameters — span, taper ratio, sweep, and twist — to maximise lift-to-drag ratio at a cruise condition using XFLR5's Vortex Lattice Method and 3D panel solver. You will script batch analyses in Python, post-process polars, and produce a documented optimum design.
Autonomous Drone Missions with ArduPilot Lua
Script your drone's brain: autonomous survey and inspection in Lua.
Use ArduPilot's built-in Lua scripting engine to write custom autonomous mission logic that goes beyond standard waypoint navigation. You will implement a grid survey pattern, an adaptive inspection hover sequence, and a geofence-triggered return-to-launch — all running onboard without an external companion computer.
Deep Learning for Engine Prognostics
Teach an LSTM to hear an engine degrading cycle by cycle.
Build LSTM and 1D-CNN models in TensorFlow/Keras to predict turbofan engine remaining useful life from multi-sensor time-series. You will design sequence windowing pipelines, compare architectures, apply learning rate scheduling and early stopping, and interpret model predictions with attention weights.
Neural ODE Orbit Propagation with JAX
Let automatic differentiation learn orbital physics from trajectory data.
Use JAX's autodiff and ODE solvers to build a Neural ODE that learns satellite orbit dynamics purely from observed position/velocity data. You will compare the learned propagator against classical two-body and J2 models, and explore how the neural model captures unmodelled perturbations.
Solve Burgers Equation with PhysicsNeMo
Train a neural network to satisfy a PDE — no simulation mesh required.
Use NVIDIA PhysicsNeMo to build a Physics-Informed Neural Network (PINN) that solves the 1D viscous Burgers equation, a canonical nonlinear PDE capturing shock formation. You will set up collocation training, enforce boundary and initial conditions as soft constraints, and compare PINN solutions against a finite-difference reference.
Design a GPS Constellation in STK
Arrange satellites in space so no point on Earth loses navigation coverage.
Use Ansys STK to design, analyse, and iterate on a navigation satellite constellation for global positioning coverage. You will define orbital parameters for a Walker or custom constellation, compute coverage metrics and PDOP maps, and trade constellation size against coverage gaps to arrive at a justified design.
Thrust Vector Control Mount in SolidWorks
Design a gimbal that steers a rocket engine — then prove it will hold.
Design a two-axis gimbal-based thrust vector control (TVC) mount for a small liquid rocket engine in SolidWorks. You will perform parametric CAD, run FEA under maximum thrust and side-load cases, iterate on weak areas, and produce a complete design package including assembly drawings and a stress analysis report.
CubeSat Flight Software with F Prime
Build the command and data handling brain of a CubeSat using NASA's own framework.
Use NASA's F Prime (F´) C++ flight software framework to implement the command and data handling (C&DH) software for a 3U CubeSat. You will design a component architecture, implement telemetry collection, command dispatch, and a simple fault management state machine, running the full system in the F Prime software-in-the-loop simulator.
Rocket Trajectory Optimization in Julia
Find the fuel-minimum path from launchpad to orbit — mathematically.
Use Julia's JuMP optimization framework with the Ipopt solver to formulate and solve a fuel-optimal rocket trajectory problem using direct collocation. You will transcribe the continuous-time optimal control problem into a nonlinear program, solve it for a single-stage rocket, and analyse how constraints on thrust and dynamic pressure shape the optimal solution.
Kaggle Turbofan Competition Pipeline
Go beyond homework: build a competition-grade predictive maintenance pipeline.
Build a full competition-grade machine learning pipeline for the NASA C-MAPSS turbofan degradation challenge on Kaggle. You will implement advanced feature engineering, train a diverse model ensemble (XGBoost, LightGBM, CatBoost, neural nets), and apply stacking to push toward leaderboard-competitive RMSE — all while maintaining reproducible experiment tracking.
Airfoil Flow Field with DeepXDE PINNs
Replace a CFD mesh with a neural network trained on physics equations.
Use DeepXDE to build a physics-informed neural network that solves the 2D steady incompressible Euler equations around a NACA 0012 airfoil, predicting pressure and velocity fields without a traditional computational mesh. You will enforce no-penetration boundary conditions on the airfoil surface and validate against panel method results.
RL Autopilot for JSBSim Flight Simulator
Train an agent to fly a real aircraft model — six degrees of freedom included.
Train a reinforcement learning agent to fly a full 6-DOF aircraft model in the JSBSim flight dynamics engine using OpenAI Gymnasium and Stable-Baselines3. You will implement a custom Gym environment wrapping JSBSim, design a shaped reward function for altitude and heading hold, and compare PPO and SAC agents on task performance.
Vision-Based Landing with PX4 + OpenCV
Guide a drone to a precise landing using only a camera and fiducial markers.
Build a precision landing system for a PX4-based drone using ArUco marker detection in OpenCV. A companion computer runs the vision pipeline, estimates the marker pose relative to the drone, and sends position setpoints to PX4 via MAVSDK offboard control — closing a visual servo loop that guides the drone to a 10 cm landing accuracy.
Train a Drone to Hover with PyTorch RL
Implement PPO from scratch and watch a quadrotor learn to balance itself.
Implement the Proximal Policy Optimisation (PPO) algorithm from scratch in PyTorch and train a quadrotor to hover and track waypoints in a lightweight physics simulation. By building PPO yourself — actor-critic networks, GAE advantage estimation, clipped surrogate loss — you gain deep understanding of why modern RL algorithms work.
Drone Simulation in NVIDIA Isaac Sim
Build a photorealistic drone test environment where perception algorithms actually work.
Use NVIDIA Isaac Sim (built on Omniverse) to create a photorealistic drone simulation environment complete with physics, lighting, and synthetic sensor data. You will attach a simulated depth camera and IMU to a quadrotor model, generate training data for a visual odometry pipeline, and validate a perception algorithm that would be difficult to test safely in the real world.
Wing Structural Analysis in Simcenter
Put a wing box through its paces with industrial FEA before it ever flies.
Perform a complete finite element analysis of a wing box structure under flight loads using Siemens Simcenter Nastran. You will build a shell-element wing box model, apply aerodynamic pressure loads from a VLM solution, run static, buckling, and fatigue analyses, and produce a certification-quality stress report.
Thermal Digital Twin with Ansys Twin Builder
Distil a full FEA thermal model into a real-time digital twin that runs on a laptop.
Build a reduced-order thermal model (ROM) of an avionics bay using Ansys Twin Builder. You will run high-fidelity FEA thermal simulations in Ansys Mechanical to generate training snapshots, apply model order reduction to create a fast ROM, deploy it in Twin Builder connected to real-time heat dissipation inputs, and validate the ROM accuracy against new FEA solutions.
Predict Composite Laminate Failure with scikit-learn
Teach a model to predict how and where a composite will fail
Generate a dataset of composite laminate configurations using classical laminate theory, then train a multi-class classifier to predict failure mode — delamination, fiber breakage, or matrix cracking — from layup parameters. Bridges textbook composites theory with practical ML.
Predict Jet Engine Thrust from Sensor Data with Random Forest
Use NASA engine data to learn what sensor readings reveal about thrust
Use the NASA C-MAPSS turbofan engine simulation dataset to build a regression model that predicts thrust output from temperature, pressure, and spool speed sensor readings. Explore feature importance and gain insight into gas turbine thermodynamics through data.
Predict Aircraft Noise Levels with Neural Networks
Use NASA wind tunnel data to predict how loud an airfoil is
Use the NASA Airfoil Self-Noise dataset from the UCI Machine Learning Repository to train a neural network that predicts sound pressure level from airfoil geometry, wind speed, and angle of attack. A clean regression problem connecting aeroacoustics with deep learning.
Predict Clear-Air Turbulence from Weather Data
Build a classifier that warns pilots about invisible rough air
Build a machine learning classifier that predicts turbulence severity (none/light/moderate/severe) from atmospheric variables using real NOAA pilot reports (PIREPs) and reanalysis data. Tackle one of aviation's most challenging weather hazards.
Detect Surface Defects in Aerospace Parts with CNN
Train a deep learning model on real industrial defect data
Train a convolutional neural network to classify images of metal surfaces as defective or non-defective using the NEU Surface Defect Database. Build a production-quality defect detection pipeline applicable to aerospace manufacturing.
Classify Aircraft Types from ADS-B Trajectories
Identify what's flying overhead from how it flies
Extract flight trajectory features from OpenSky Network ADS-B data — climb rate, speed profile, turn radius, acceleration patterns — and train a classifier to identify aircraft types. Learn feature engineering on spatiotemporal data.
Predict Flight Fuel Burn from Route and Aircraft Data
Build an ML model that estimates fuel consumption before takeoff
Use publicly available flight data from BTS or Eurocontrol to build a regression model predicting fuel consumption from distance, aircraft type, payload, and weather. A real-world ML problem with direct sustainability applications.
Analyze Aviation Incident Patterns with NLP
Apply modern NLP to uncover hidden patterns in safety data
Use the NASA ASRS database to classify incident narratives by category using TF-IDF and transformer models. Discover patterns in human-factors incidents that traditional analysis methods miss.
Predict Wing Flutter Speed with Gradient Boosting
Train XGBoost to replace hours of aeroelastic simulation
Build a dataset of wing configurations with flutter speeds from analytical or simulation results. Train a gradient-boosted model to predict flutter onset — creating an ML surrogate for computationally expensive aeroelastic analysis.
Predict Avionics Bay Temperature with Regression Models
Build an ML model for the thermal challenge every aircraft faces
Build a regression model predicting peak avionics bay temperature from flight phase, ambient conditions, and power dissipation. Use synthetic data from thermal network models to train and validate your approach.
Other Levels
High School
Hands-on projects for students with basic math and science. No prior coding experience required — every project walks you through setup from scratch.
36 projects →Advanced
Research-grade projects for grad students and ambitious undergrads. These involve cutting-edge techniques like PINNs, reinforcement learning, and multi-sensor fusion.
32 projects →