Term 2 is moving along fairly quickly in comparison to Term 1. I just completed the third project (Kidnapped Vehicle Project) and will most likely wrap up the term mid-December. It’s interesting that the pace has slowed – although I would say the content is a bit more strenuous in terms of what we are learning. That possibly could be due to using C++ (my first time completely projects in the language) though.
The past two projects (Kidnapped Vehicle Project and Unscented Kalman Filter) build on the use of LIDAR or Radar data to track a car. While the concepts were fairly difficult to grasp (I’m still trying to wrap my head around Particle Filters and Kalman Filters), the implementations were very rudimentary. By no means would our implementations be used in a true self-driving car and I would love to have more exposure to code that is being deployed on Waymo, Cruise, or Uber ATG vehicles. As a side note, you can go check out my code for the Kidnapped Vehicle Project here and the Unscented Kalman Filter here.
Looking forward, we’ll be moving into control (versus localization) in the next project. My guess based on seeing small previews is that it is similar to the Behavioral Cloning project (minus the deep learning). We’ll most likely be learning how to control steering angle and throttle responses given a virtual track outline. What would be most interesting to me, would be to learn how to tap into modern cars via the can bus. Quick introduction to the topic can be found here. I doubt this will be the topic of the lecture and project, but it definitely gives me some extracurricular work to pursue after I wrap up the Nanodegree.
As I work my way through Term 2, I’ve found myself wanting to dive further into the Artificial Intelligence side of computer programming. Specifically, building GAN networks seems like a fun exercise. Giving some reflection on this course as a whole, I think it’s best to treat it as an introduction to the technologies that are being utilized to control a self-driving car. With this mindset, you can then dive further into a subject (sensor fusion, deep learning, etc.) and pursue this track on your own time OR pursue a Master’s or PhD degree.
Additionally, I believe this will increase your chances for landing a job in the self-driving space (or even another sector: finance, drones, etc.). Many of the Udacity students have complained or voiced the opinion that many employers don’t believe the course load of the Self-Driving Car Nanodegree is nearly rigorous enough to prepare you for a self-driving car engineer role. While some students have received and pursued offers, they have typically had undergrad CS, EE, or similar degrees and exemplified a different level of expertise than other students.
For those who are just being introduced to these topics, I definitely recommend treating this as an introduction to the space. Put some time and thought into what interests you most from the Nanodegree, and spend time building/tinkering/contributing to open source projects. Then, possibly even pursue higher education. If you are looking to save some money and can’t go to school full time, check out the Georgia Tech OMS CS. At less than $7,000 total, you can’t really beat the price. Plus, reviews of the degree say that it is more than worth it.
I’ll definitely be putting some thought into my next move once the Nanodegree wraps up around March. The big question mark for me is whether to pursue an MS CS or MBA. I originally pursued programming to become a better Product Manager and an MBA can be useful on this front. But, if I want to pursue the self-driving car space, I will most likely need to go for an MS CS. This will most likely even be true for being a PM in the self-driving car space due to the deep understanding of the technologies at use. I’ll be writing more about this in the coming months 🙂