Learning based localization for indoor positioning systems

Finding and tracking position indoors

Localization has been playing a major role not only in navigation, but also in areas such as security, surveying, timing, and it is now difficult to imagine living without it. GPS like systems allowed incredible satellite-navigation capabilities to be accessible in almost everyone’s pocket. However, despite its various accomplishments, GPS systems need direct line of sight, making it nearly impossible to pinpoint accurate locations indoors.

My dissertion work tackles this problem using RSS localization methods, in particular due, to its cost, accuracy and scalability. RSS-based tecniques use a statistical model based on the power of transmission and path loss exponents (PLE) to compute the distance and estimate a solution. Even though there are already algorithms that solve this problem applying optimization based methods, such as maximum likelihood estimation and convex relation techniques, even when channel parameters are not known, all of these methods assume PLE is constant in the same propagation medium. However, practical experiments showed signal path loss can take unexpected values, mainly due to multipath and shadowing. Therefore, a robust enough model to handle these changes is needed.

The work I'm developing envisions obtaining an algorithm to achieve an efficient solution by tuning and enhancing PLE estimations, using optimization and machine learning procedures.

Practise of Large Scale Machine Learning

A 7-day intensive course in a top institution in Paris

Twice a year, the ATHENS Network (Advanced Technology Higher Education Network / SOCRATES) organises a programme of 7-day intensive courses at member institutions. The programme known as the ATHENS week includes a wide range of courses in engineering and social sciences as well as a number of “European Dimension” activities in each country. This Network was created in 1997 is made up of 15 European technological universities and nine ParisTech Graduate Schools of Engineering, which coordinates the programme.

I attended the course "Practise of Large Scale Machine Learning" in Télécom ParisTech, which allowed me to:

• Learn and consolidate previous knowledge of supervised machine learning, especially deep learning in neural networks;
• Practise with large datasets using 3 different frameworks, such as PyTorch, TensorFlow and Keras;
• Get familiar with jupyter notebook;
• Network and share experiences with students all around Europe;
• Experience and live abroad in one of the world's biggest cities.

• Date: November 2019

Smart Gym

The management of a project in JUNITEC.

This project was proposed by one major chain of gymnasiums, when I was working in JUNITEC. It consisted in the creation of a monitoring network of fitness machines, in order to evaluate the intensity of use. The network was made to cover more than 100 machines per gymnasium and presented multiple challenges during the projected, whether in the choice and production of the sensors itself to the connection of all the devices in an IoT network, that could be accessed through an online platform.

As the project manager, my tasks included:

• Planning and defining the timelines of the project
• Schedule weekly meetings and divide tasks with the team;
• Constant communication with the client;
• Report and follow-up each stage of the project.

University Projects

Find more about the college projects that gave me important tools

During the 5 years of my integrated Master's Degree in Electrical and Computer Engineering, I frequently had to develop projects for the various courses I completed, as a way to consolidate what I had learnt. Most times, I had to face lots of challenges to deliver a solution. My faculty never made it easy for me, I always had to ask lots of questions and discuss with my colleagues in order to put the pieces together. As I like to say, they teach us what is an engine, the wheels, the gear box, etc, and we have to figure out on our own how to make a car.

European Parliament - Euroscola

Taking the role of decision-makers in Strasbourg, France.

The European Parliament set up the Euroscola project in 1990, thus allowing hundreds of thousands of high-school students to participate in a simulation exercise of the work of MEPs.

The participation was awarded to my school, after making it through a national competition organized by the European Parliament Information Office. After preparing the issues previously established, we came to Strasbourg, where we got together with 500 students to debate, take sides, negotiate, amend, vote and finally adopt resolutions on real European issues.

This experience allowed me to familiarize with the workings of the European institutions, to discuss democracy, fundamental rights, European values and to express my personal opinions on the decisions taken at the European Union level, by experiencing it firsthand. It was also an opportunity to practise my language skills and making friends with fellow students from across Europe.

• Date: November 2012

Fina Olympic Marathon Swim Qualifier 2012

An international voluntary work experience

Open water swimming competition in which 101 athletes from 44 countries took part. It was the last opportunity to qualify for the London Olympic Games, in Setúbal, Portugal. Only 9 men and 9 women booked a place for the Olympic marathon swimming race in Hyde Park, London (August 9-10).

As a volunteer, my tasks included:

• Accompaniment of athletes and coaches from the airport to the hotel;
• Monitoring the swimming practises;
• Overseeing the athletes in the anti-doping process;
• Secretarial services and installation of facility where the competition took place.

In addition to get to know various professional athletes and their coaches, I had the luck of being responsible for overseeing the athlete who ended up winning the gold medal in London, Oussama Mellouli.

• Date: June 2012