About Simulation Techniques

The final project of my undergraduate degree, which was entitled "Multicast Switching Technologies in High Speed Routers", was supervised by Associate Professor Fan Yang. I chose him to be my supervisor as I had a very positive impression of him in his lectures on switching technologies. One day I sent him an email to discuss a statement from the textbook, which had made me rather confused. "Your understanding is correct, and the way to express the idea in the textbook is a little misleading," he responded quickly. "How would you like to do the project?" he asked me. "If you really want to do an excellent project, I can guide you to use a couple of network simulation tools such as NS2 or OPNET." "Yes, I want to try my best and learn as much as possible from it!" I said with a determined voice.

However, I soon became regretful for this. In the final semester of university, I first devoted myself to finding a job. I wanted to find an employer who could allow me to break the contract with ZTE. After many unsuccessful attempts, I had given up. Instead, I used almost all available time to prepare for taking the IELTS test, with a plan to study in Australia. After finishing the examination, it was merely a week or so before the due date of the final project. At the time, it had no longer been enough for me to learn those simulation tools. I was thinking of writing a program by myself from scratch. I wrote an email to my supervisor about the basic idea, and he confirmed with me that the basic idea was rational. Therefore, in my final project report, there were a number of pages of a C program appended, which simulated an algorithm from a paper. It was a console program without any visualization. What the program did was simulate the algorithm to calculate throughputs.

In July of 2008, when I'd just graduated, I was engaged in the training program for fresh graduates in ZTE. I was assigned to the department of CDMA network planning and optimization. In the real world, network planning typically relies on simulation as the initial step. This is because setting up a wireless network is of high cost and complicated. Many calculations are done inside the simulation tools. After the installation of network facilities, a road test is compulsory to test the signal strength in every corner of the place, and adjustments are usually needed as a later step.

In 2009, I enrolled in a course at QUT in Australia. In the second semester, I enrolled in a unit about operating systems. The virtual machine is a type of simulation which simulates the hardware from the ground up. At the same time, I was doing a project in the Microsoft eResearch Centre. The project was related to the Windows CE embedded system, where simulation was used as a tool for development.

In early 2010, I enrolled in a unit called "AI for Games", which taught computer games agent development. This was probably my favourite unit at QUT. In this unit, we explored the OpenSim platform, an open-source project written in Mono.NET. It was a Second Life simulation platform. Various algorithms had been learned during this time, with the honey bees optimization algorithm, the Mini-Max human chess competition algorithm, and the distributed behaviour models that used to model collective behaviours among the most impressive ones.

Several years later, I began to draft a thesis on the topic of "Unusual Event Detection in Crowded Scenes" in the area of intelligent computer vision. I realized that many abnormal events with security interest are impossible to obtain with a sufficient number of instances in real-world surveillance footage. "When we can't get them in the real world, how do we test them?" I wondered. Then I suddenly thought of the OpenSim simulation that I'd learned before. If the videos generated by computer graphical simulation were not looked as real enough, I thought we could also use a set of robots to simulate. While a group of robots moved together, we could apply the distributed behavior models to simulate movements in the real world. Then in the conclusion chapter, I said cross-disciplinary research, such as including simulation techniques, was compulsory in this research topic.

In 2015, I was working on a paper on resource allocation management in the area of process mining. We first used simulated data with a well-known ground truth to evaluate the algorithm in ideal conditions. Then we applied it to a real-world dataset, and verified it with business feedback. Professor Wil van der Aalst, who was the "father of process mining", had given the section I drafted very positive comments, which I had been very proud of until now.

However, I do believe simulation is overused in telecommunications engineering research. I've read many papers in these areas, and many of them only have simulation-based results. I've always considered those papers as incomplete. "When are you going to start simulation?" Professor Huang urged me when he was pushing me to have a paper in pattern recognition, where simulation was rarely used, when I was working in Shenzhen University.

In around 2020, I started some development for mobile applications. Simulations for different models of mobile devices were used extensively to develop mobile apps on a desktop computer.