How to Win: Conversations with MathWorks Minidrone Competition Winners
Today, we will talk to various teams from around the world who participated in the MathWorks Minidrone Competition! They will share their thoughts on their experience about working with Simulink through the MathWorks Minidrone Competition.
Before we dive into our conversations with the teams, I would like to give a quick background about the MathWorks Minidrone Competition. The MathWorks Minidrone Competition is a platform for students to learn model-based design using Simulink. The competition typically consists of two rounds: Simulation Round and Deployment Round. The participating teams design a minidrone line follower. The top 5 to 7 teams whose algorithm performs the best in simulation are shortlisted for the Deployment Round. Learn more here.
Pranav Murali, Team Beeclust, SRM Institute of Technology
We will talk to Pranav Srinivas Murali first; Pranav was part of a research team called Beeclust Multi Robot Systems Lab at SRMIST India. His team participated and won second place at the MathWorks Minidrone Competition held at IROS 2019 Conference. He also joined MathWorks for an internship at the new graduate program after the competition.
How was your experience learning Model-Based Design through the competition?
Model-Based Design helped our team intuitively understand all the sub-systems within the drone. Broadly, the competition required image processing, path planning, and flight control blocks to realize the problem statement. We gained vital insight on the features and capabilities of each block to constructively develop our algorithm. Without the Model-Based Design, it would have been harder to calibrate and tune the various parameters required to achieve a perfect flight. I would certainly agree that model-based design is a great approach to divide and conquer complicated systems.
Was there anything new you learned from the competition?
Actually, this was the first time our team worked on a drone. We sat together and designed the approach that we would have the drone follow and designed the simulation. But when we ran the simulation, we were exposed to the flaws of what we thought was a strong algorithm. The drone failed to follow certain tracks. We then designed multiple approaches to back up the algorithm for unforeseen issues. This provided us a chance to learn all the possible ways to model our system.
For the final round, we had to deploy the algorithms on the drone. Here, we learnt that we should consider for all real-world parameters in the algorithm since they were different from all the assumptions we had in simulations. The whole event was indeed a fantastic learning experience. Post the competition, all the participating teams discussed their algorithms which proved that learning is a continuous process.
What are your thoughts about the MathWorks Minidrone Competition?
The MathWorks Minidrone Competition is an excellent platform for students looking to get a hands-on experience with the Signal, Image Processing, and Control Systems concepts taught in the classroom. The concepts are understood better when they are seen in action. At our lab, the line follower robot is the first project a fresher makes, since it is one of the simplest links between Robotics, Hardware, and Software. This competition provides not just that, but involving a drone. The problem statement can be solved in multiple ways, by changing the algorithm used under each block. So the solution is only limited by one’s creativity.
Dr. Diana Marcela Ovalle Martínez, Assistant Professor, Universidad Distrital “Francisco José de Caldas”, Colombia
Next, we talk to Dr. Diana Marcela Ovalle Martínez who is Assistant Professor at Universidad Distrital “Francisco José de Caldas” in Colombia. Her students won second place at the MathWorks Minidrone Competition at Colombia 2019 and participated in the competitions held at IROS 2019, Macau, and MATLAB EXPO UK 2019.
How was the experience to participate in the MathWorks Minidrone Competitions?
I remember I saw for the first-time publicity related to the competition on Facebook through the MATLAB & Simulink Robotics Arena Facebook Group. The competition experience was new for all of us. We got to participate in three different MathWorks Minidrone Competitions, noticing differences mainly in teams that participated and the arena, which made every competition a unique experience for the team. Prior experience about the competition can possibly help to perform better at the next one, but anxiety in the competitive environment can play against you.
At times we did get frustrated with the initial results, but then we always thought about how to improve and, eventually, had the satisfaction of seeing our vehicle follow a line, land, and completing the entire circuit.
How do you think Simulink may help you in your future projects?
Simulink is a complete tool for simulation. If you continue in the line of control and robotics, project development will always be supported by Simulink, not only to verify the dynamic models but also to keep in mind the perspective of model-based design. Right now, with Covid-19 contingency and remote teaching, Simulink has helped me to put together nonlinear simulation models for different systems in order to help my students design controllers and, even, to recreate robot’s behavior in virtual environments.
How do you think this will help your students learn better?
As an engineer, I am a believer that we as professors need to be motivating. I think developments in virtual environments help a lot in visualization, because it is always nicer to see a vehicle in motion than a system signal in a Scope. So I am already including this component in simulations to motivate students to design better controllers, and it is working very well for us. Also, Stateflow is a convenient tool for behavior-based robot control, which will be also a nice addition to my course’s simulations repertoire.
The use of many new capabilities of Simulink, since they are quite straightforward to use, allow the student to focus on solving the problem than in the commands syntax to make the code work.
These competitions really help us, professors, to update our knowledge of such amazing tools developed over MATLAB and Simulink.
Yuqiu Yu, Team SJTU Aerovision, Shanghai Jiaotong University, China
Next, we will listen in to Yuqiu Yu, team leader of SJTU Aerovision, from Shanghai Jiaotong University, China who won the MATLAB Award at the Minidrone Competition held at SYSU.
How did you know about the Minidrone Competition?
It was our advisor, Prof. Shiqiang Hu, who introduced us to this competition. We also heard about the competition from the WeChat account. Prof. Hu suggested that we take the opportunity of this competition to apply what we have learned on a fast and high-intensity stage, communicate with students in various colleges and universities, and get benefits for our research.
Was there anything new you learned from the competition?
We learned a lot from the competition. This was our first time using Simulink. And we used Simulink to quickly build a complete autonomous flight system to include sensor data preprocessing, image processing, state estimation, and perception. And finally deployed a system to the actual Minidrone. The best part was having the flexibility to modify the model parameters to see the actual flight effect and monitor the flight data. The workflow greatly improved our work efficiency.
What were the challenges and how did you tackle them?
Our biggest challenge was the computational cost. In the simulation round, the computing platform was a laptop. Although our algorithm was relatively complex, it could be simulated at a higher frequency on the laptop, so the simulation results were very good. What we missed to understand was that this code would be deployed on a drone with a limited computational power. The operation frequency of the algorithm is very low on the drone, resulting in the failure of the minidrone to track the line.
To tackle this, we used the Monitor and Tune mode on the drone to see the values from the hardware live as they update. We changed all the blocks that may have had large computational overhead on our drone, and were super happy to see the drone fly as it was expected to.
Shane Malone, Team UCD, University College Dublin, Republic of Ireland
We now will hear from Shane Malone, team leader of Team UCD from University College Dublin, who won first place at the MathWorks Minidrone Competition at Cambridge 2020, one of the competitions held virtually this year.
How was your experience learning Model-Based Design through the competition?
This challenge marked our first real foray into model-based design and with that came a steep learning curve. We worked our way through the excellent support and Onramps provided by MathWorks, and saw how MATLAB, Simulink, and Stateflow could be combined to perform powerful operations. At the start, the model to get started intimidated us. But visual environment and clean layout of Simulink made the process of visualizing our algorithm easier, allowing us to quickly iterate design ideas.
This was our first time using Simulink and it will definitely not be the last. We have seen the powerful operations that can be implemented in Simulink remain readable and easy to follow for the user. We intend to continue using Simulink in our projects in University and in some side projects.
How was the experience working virtually on the competition problem statement?
This competition, like many things, was initially meant to take place in person but had to move to an online platform due to the COVID-19 pandemic. Despite this disappointing revelation, the team at MathWorks worked exceptionally hard to put on a remote event and it was a great opportunity to see how other teams had approached the problem. While we could not be together in person, it was an enjoyable day and event.
Working virtually as a team had its complications and we no doubt would have preferred to be together. However, we managed the workload evenly, took initiative to work away on separate aspects of the design and checked in often through webchats to ensure we were on track. The project would have been near impossible if it were not for us learning how to use Git to control versions of the model. This was helped by the Git integration built into MATLAB.
Any recommendations about the competition to the readers?
My advice to anyone considering taking part in this competition is to just go for it. While it may seem daunting and overwhelming at times, there is nothing more satisfying than seeing the end result of your hard work. It is a great opportunity to learn more about image processing, path planning, model-based design and more. There is a plethora of resources available from MathWorks as well so you will have plenty of help. So, go on, take on the challenge!
Meet all the MathWorks Minidrone Masters who are the winners of the most recent MathWorks Minidrone Competitions. And, if you would want to host something similar on your campus, club, or class, drop us an e-mail at minidronecompetition@mathworks.com.
评论
要发表评论,请点击 此处 登录到您的 MathWorks 帐户或创建一个新帐户。