Simulating autonomous airplanes!

1) Automated data reviews using Bayesian statistics to create confidence intervals using historic test data saving 50 labour hours/week.

2) Used OpenCV to scan for defects on solar cell EL images and built a PyQt5 app to make go/no-go decisions using historical data statistics.

3) Integrated internal systems in a chain of 27 SQL, HTTPS, and API queries along with low-level vendor software to automate end-of-line

ATP testing leading to complete elimination of human error and 10% reduction in labor hours.

4) Created data models to track and predict labour hours across two production floors using SQL and PowerBI.

5) Designed PID loops in successive loop closure using ladder logic to tune a 3-axis CNC machine resulting in +/- 0.5mm^2 precision in end-

effector placement.

• Designed an autopilot with a strong successive loop closure algorithm at its core in hopes to create a versatile controller that can be

tuned to the vehicle constants of almost any RC aircraft.

• Calculated the vehicle physical constants (lift and drag coefficients) of our test RC aircraft by processing raw test flight data through a

converging aerodynamics model created using the AVL software.

• Created a Gaussian Wind Model using a simple gaussian generator for estimating the time and intensity of the wind bursts.

• Created a simulator using PyQt5 to assess the autopilot’s effectiveness with true and estimated states along with added wind and real-

time human controls.

Creating robust guidance systems for the purpose of stochastic motion planning in uncertain environments with the goal of gaining maximum information about said unknown environment.

• Mapping out flavours and filters of the original RRT algorithm using MATLAB

• Simulating solutions in different environments using Gazebo and ROS architectures.

• Processing using Jetson Nano's Linux architecture for simulating real-flight computation power and runtimes.

Visual SLAM techniques such as ORB SLAM as well as waypoints are used for localisation and flight controllers such as the PixHawk for control. These, along with our guidance algorithms provide the full GNC for the flight article.