It's been a year since I started MSc in control engineering. I studied a lot of topics and the most topics that attracted me was related to trying to implement the control algorithms like PID in a micro-controller. My last project was an aerial quadcopter, and it was a big one. However, I didn't like its control algorithm (PID) and was searching for another control algorithm to use it but it seems very hard in my current state. So I was hoping if you can help me with new topic that I can implement it later for my drone. I was thinking about machine learning, but don't know from where should I start if my goal is control.
If the PID controller works fine for the flight control, there is no reason to change it. In my opinion, the most important control challenges for your application is at supervisory control level: Collision advoidance, mission tracking, fault detection, fault tolerant control (i.e. what can be done if a motor fails), control system reconfiguation, video camera feedback for fulfillment of a mission, flight optimization for saving battery. I think you have many many thigs to do. It depends on you interest and motivation.
I was working on a hexacopter on Collision avoidance and mission tracking, but as i read before, if i want to write a paper, anything related to this topics is not good. That's why i am trying to focus on something new i can add to the control part. PID won't be good if the weather is windy. And i am looking more about a topic that can help me after finishing my MSc to get a job.
If you would like to write a paper, you need to produce a new contribution. You were building up your background and the experience in control as well as in the application. Now you need a research topic, in which you can contribute with something new. If you said the PID control is not good in windy days, then you know the problem, you need a new solution, i.e. your new solution, it does not matter which algorithm. Have a look at the literature, if you find a good solution you can implement it but is is not yours and then it is not clear if you are producing a new contribution to the field. But the way, do not thing big but step by step. Machine lerning could be a subject, but you will need a very long time to obtain new results different from literature. Your system is nonlinear and stochastic (in the presence of wind), maybe adaptive stochastic control (see Astrom's Adaptive Control book) could be an alternative to the PID.
- Badges
- Science topic