14 - Seminar: Optimization of Traffic Signal Control using Machine Learning

Hi crew,

Are you alright?

Today I want to talk with you one more time about Machine Learning (ML). In fact, last week I attended an interesting seminar about possible applications of these techniques. The seminar has been given by the TRL Academy, here in TRL, and for me, it was a good opportunity to introduce myself, my project, and to share ideas about possible optimization algorithms and techniques, for a topic which is completely new for road asset management.
In my previous posts, I showed you how ML works, giving particular attention to Artificial Neural Networks and Random Forests. As I already mentioned, these techniques are used nowadays in a number of applications in Science, and actually, also lots of apps installed on our mobiles use similar methods to make our life easier. In fact, based on the ability of these techniques to recognize patterns in large datasets, weather forecasts, banks, social media, among many others, use ML to improve precision and reliability of the results obtained using a classical approach.

Fig. 1 – Scheme representing the idea of Machine Learning (WordPress.com)
TRL is a visionary company in this sense and they are trying to apply these techniques to a number of studies. In particular, here in TRL, the idea is to use these techniques to improve the current road asset management strategies. Therefore they got interested also in my project. The seminar that I attended last week is just an example of the enormous potential of ML applications.
The study carried out and presented by Dr. Islam Abdelhalim, Anatoly Smirnov from the Software Group, and Rahul Khatry from the Safety and Technology TRL Department shows how Neural Networks, Decision Trees, and other ML techniques can be successfully used to optimize the traffic signal control for a case study in the Netherlands.
As you can see, crew, I am not the only crazy person trying to understand and apply Machine Learning to research. Apparently, there are lots of crazy people around the world. Hehe

Fig 2. – Scheme of the optimization process for traffic signal control presented in TRL;

In the study conducted in TRL a combination of Decision Trees and Neural Networks has been used to optimize the timing of traffic lights based on real-time data about the traffic flow around the considered area.
It was really interesting, but not surprising (for me), to see how the Neural Network was able to adapt its behaviour to the training dataset proposing an optimized timing pattern for the traffic lights very close to what the previous platform was suggesting. The Network was able to solve some issues of the older methodology, improving in some cases the overall performance of the system reducing also the processing time. And this is just for a case study. Try to figure out what can be the advantages when thousands of traffic lights need to be analysed. Trust me, it can produce a huge improvement.
So, this is the power of Machine Learning. And this are some of the reasons why I’m using these techniques in my research project. Cool, eh?

In the end of this post I would like to thank Dr Helen Viner to inform me about this seminar, and to Dr Islam Abdelhalim, Rahul Khatry, and Anatoly Smirnov, for the nice discussion we had at the end of their presentation regarding the use of Neural Networks in road asset management applications. It was a great opportunity for me to learn and share ideas about possible methods and techniques to use in my project.
Finally, one more thing guys: I applied to the 4th MSCA prize 2017. It is a prestigious prize given to people involved in Marie Skłodowska-Curie actions for their contribution to Science. Please crew, keep finger crossed for me. You know, it would be an honour for me to be awarded with such a prize.

Thank you all for your attention and support.
Hoping that this post was of your interest, as usual: stay tuned!




Popular posts from this blog

28 - Updates on progress of my project

13 - My secondment in TRL

27 - Road to Project Management