9 October, 2026 at 9:00 AM
Anum Umer, "Integrated Sensing and User Localization in Reconfigurable Intelligent Surface-Assisted Dynamic Rich Scattering Environments"

Supervisor: Professor Muhammad Mahtab Alam, Thomas Johann Seebeck Department of Electronics, School of Information Technologies, Tallinn University of Technology,
Tallinn, Estonia

Co-supervisor: Dr. Ivo Müürsepp, PhD, Thomas Johann Seebeck Department of Electronics, School of Information Technologies, Tallinn University of Technology, Tallinn, Estonia

Opponents: 

• Professor Luca Reggiani, Dept. of Electronics, Information and Bioengineering, Politecnico di Milano, Milan, Italy
• Professor Riku Jäntti, School of Electrical Engineering, Aalto University, Espoo, Finland

Wireless networks are moving towards the sixth generation known as 6G. One key goal of 6G is to find the position of a device with very high accuracy. This matters indoors where satellite navigation is not available. Typical examples are industrial facilities and warehouses and smart buildings. A reconfigurable intelligent surface is a flat panel built from many small electronic elements. Each element can be tuned electronically. The panel can then reshape radio signals and steer them where they are needed. It works like a smart mirror for radio waves. It is low in cost and uses little power. It can also carry signals into areas that the base station cannot reach directly. Some of its elements can sense the surroundings at the same time. Indoor spaces are full of objects that reflect radio signals. People and machines move about and change the signals all the time. This makes the position calculation very difficult. Most existing methods assume that the surroundings stay still. Their accuracy drops when the environment keeps changing.
This thesis studies how these panels can be used for positioning in such changing environments. It makes four contributions.
The first contribution is a broad review of published research on this topic. It sorts the field into five dimensions and points out the open problems.
The second contribution is a method that selects a panel setting from a list of settings learned in advance. The choice is guided by what the sensing elements observe about the moving objects. This reduces the position error by up to 97 percent when compared with random settings.
The third contribution is a method that needs no prepared list at all. It estimates where the moving objects are and then designs the next panel setting step by step. It reduces the error by up to 79 percent as the panel grows from 20 to 100 elements. It also reaches the same accuracy using about half the number of test signals. This cuts the time needed for positioning by half.
The fourth contribution extends the method so that the system finds both the position and the facing direction of the device. The panel setting and the antenna beams of the base station and of the device are designed together. A mathematical benchmark is also derived for comparison. The learned method performs better than this benchmark because it plans over the whole sequence of measurements rather than one step at a time.
The results were produced by computer simulation. Taken together the work shows that positioning with these panels is best treated as a step by step learning process that is aware of the moving surroundings rather than as a fixed geometric calculation. This offers a practical path for industrial and warehouse and smart building deployments.

Join the public defence in MS Teams

Meeting ID: 310 209 889 671 751 
Passcode: Ts6RR95J