The Flying Car Nanodegree Term 1
The real deal.
Here’s a 50 feet view of the Complete Course:
The first term consists of seventeen lectures and four projects. Personally, I’m a fan of the Production quality. I think they are the MKBHDs of Education when it comes to the Video Production.
MATH! The Nanodegree is very very Math Heavy. If you’re a Math Person, you’d be much more comfortable on here.
The Programming too, is very challenging.
Udacity mentions the Nanodegree to be Advanced, I believe that is the toughest Nanodegree. The Code includes implementing Algorithms, and Math! and much learning of about physics. I really wish I had been a better student during school.
- Introduction: Introduction to Autonomous Flight and the Basics.
- Path planning: Path planning introduction and beyond: from A* on a grid to a real-world 3D planning.
- Control: Vehicle dynamics, control architecture(cascade PID controller) and full 3D control.
- Sensor Fusion and Estimation: Introduction to estimation, sensors(GPS, IMU, etc.) and extended/unscented Kalman filter.
Link to My Github: https://github.com/init27/
Backyard Flyer Project
We used Event Driven programming to create a basic Flight Controller.
This was very basic, I assumed the ND is a piece of cake- I was very wrong.
We start from A* Search Algorithm in the 2D world to 3D motion planning using random sampling, graphs.
This project is a continuation of the Backyard Flyer project, we coded path planning through an urban environment in 3D!
The Math in these lectures literally made gave me nightmares! This part had the toughtest lessons and Project.
The lectures start with vehicle dynamics.
The Key things taught were:
– PID Controller.
– Cascade PID Controller applied to a Drone.
This lecture was so difficult that they eventually relaxed the requirements and made us submit just the C++ implementation and removed the need for the Python submission.
The project consists of the implementation and tuning of a cascade PID controller for drone trajectory tracking.
Sensor fusion and Estimation
This was divided into:
- Intro to Estimation.
- Kalman Filters.
We develop the estimation portion of the controller used in the CPP simulator.
There are various scenarios that we had to pass:
– Implement a better rate gyro attitude integration scheme
– The prediction step should include a correct calculation of the Rgb prime matrix, and a proper update of the state covariance.
– Implement the magnetometer update.
– Implement the GPS update.