This page is dedicated to my Master's Thesis project. The aim of the project is to apply Machine Learning algorithms to enhance the development and usage of a lightweight Blended Wing Body (BWB) Unmanned Aerial Vehicle (UAV).
The subtasks of the project can be listed as following:
- Developing an Evolutionary Multi-Objective Bayesian Optimization system.
- Create a Pareto non-dominated vector of Bezier curve based airfoils.
- Use the non-dominated airfoils in coeffect with the bodies global parameters (e.g span, taper and sweep) to find a global approximate optimal solution with a Panel Method based system.
- Find a local optimal solution using OpenFOAM/SU2 based Computational Fluid Dynamics softare.
- Test optimized wing.
- Parametrize propulsion and control surface system to project requirements.
- Set up Raspberry Pi based autopilot system.
- Test autopilot.
- Set up a Deep Q Learning (or other) based reinforcement training policy.
- Initial data collection of operator controlling vertical hovering and VTOL of the BWB with a RC control.
- Controlled environment training of vertical hovering and VTOL.
- Outdoor environment training of VTOL and tests.
- Set up camera and telemetry system for BB UAV.
- Set up software to recognize a 2m x 2m blanket as target for landing or package drop.
-
Set up communications with EPOC EMOTIV
-
Set up neural data transformation pipeline
-
Benchmark classification algorithms with crossvalidation:
a) Hidden Markov Models (HMM)
b) Ensemble Methods (EM)
c) Multilayer Perceptrons (MLP)
-
Choose and combine (if necessary) classification algorithms
-
Test BCI system
-
Set up communications with BWB UAV
-
Set up transformation and classification pipeline on UAV onboard computer
Progress so far:
For instructionss on how to set up an eGPU eGPU on Dell.