With a focus on student robotics competitions, Connell D'Souza and Kris Fedorenko show you how to get started with black box modeling. You will be exposed to basic modeling concepts, and see a demonstration of the system identification process on real-life data from Blue Robotics. Having a model allows you to design and test a controller as well as model a larger system—important steps for preparing your robot for a competition. You can find all the data used in this video on MATLAB Central’s File Exchange.
The System Identification app enables you to perform all stages of modeling such as importing and preprocessing the data, trying out different model structures, and evaluating the resulting models. Two datasets of input and output data for a T200 Blue Robotics Thruster are used to demonstrate the modeling process. Connell and Kris show how to process the data by removing means and filtering out noise. Several models are then created using simple linear model structures like state space model and transfer function, illustrating that modeling is an iterative process. You will also learn how to use validation data to evaluate your models.
After this video, you should be able to create a reasonable model for your own hardware component. As Connell and Kris underline, you would need to collect good-quality input and output data for both estimation and validation, start with simpler model structures, and keep iterating until you achieve good results. You might find the following links helpful:
