Simulink Student Challenge Winners
MathWorks is happy to announce the winners of the 2021 Simulink Student Challenge. Congratulations to the winners and thank you to all the students who showed off your impressive projects!
Reinforcement Learning Control for Wheeled Self-balancing Systems
Delhi Technological University – Kanishk K
This project shows an interesting approach to developing a controller to a nonlinear system. Kanish explained that, despite this system being widely studied, most control schemes historically relied on linearization, and that he wanted to use a different approach to better capture the nonlinearities. Modelling the system in Simulink using Simscape Multibody, Kanish implemented a Reinforcement Learning (RL) based controller. Two different RL algorithms are compared, Deep Deterministic Policy Gradient (DDPG) and Proximal Policy Optimization (PPO). Through animation and data visualization, Kanish shows that DDPG is the better choice, being better able to achieve the control objective and using less power, despite a shorter time step being used with PPO. Overall, this project is a great example of how different domains and technologies can be combined inside of Simulink to develop powerful solutions to complex engineering problems.
Dynamic Equivalent Modeling of a Black-Box Microgrid
Shenyang University of Technology – Zizhao Wang
This project models a microgrid system in Simulink which is used to generate data for training a long short-term memory (LSTM) neural network. The network fits a nonlinear model that predicts the transient behavior of the micro-grid and helps with analyzing load stability. This project interfaces with MATLAB and utilizes Simscape Electrical blocks to model various components of the micro-grid. Overall, this project is a great data-driven example of estimating parameters for an unknown complex system and has direct real-world applications.
Simulating the insulin-glucose dynamic model for Type 1 diabetes
Georgia Institute of Technology – Alexandra Zamitalo
This project demonstrates how Simulink can be used to simulate a Type 1 diabetic patient’s blood-glucose and blood-insulin levels, which can be used to determine the best insulin dosage strategy for a given patient. Alexandra models a patient’s glucose and insulin levels over time with a set of differential equations, which are solved in Simulink. Using this model, the effects of factors such as the timing of meals and insulin injections, the brand and dosage ‘type’ (short or long-term) of the prescribed insulin, and the body’s production of epinephrine can be examined via plots of the glucose and insulin levels over the course of a day. Using these plots, a proper dosing strategy can be developed for a given patient to prevent them from reaching excessively high or low blood-glucose levels. Finally, Alexandra discusses how future work with this model can involve an app to determine the optimal dosing strategy for the patient based on the patient’s eating habits and lifestyle. Overall, this is an excellent showcase of how Simulink can be used in the biomedical field to model biological processes.