Positioning Systems

Positioning and mapping applications with signals such as wi-fi or image-based algorithms in indoor areas / use of machine learning methods in positioning systems from satellite images or signals.

Reinforcement Learning Algorithms

Reinforcement Learning theory and algorithm development / Evolutionary Reinforcement Learning Algorithms / Efficient and safe discovery algorithms in Reinforcement Learning Agents.

Openstack Framework

Efficient and smart data center projects using machine learning methods through the openstack open source project.

Mission Control Algorithms

Designing deterministic or machine learning based mission control and swarm algorithms in simulation environments such as Gazebo and Unity.

Compiler Optimization

Machine learning methods on compiler optimization techniques.

Generative AI

Methods of producing music and video using generative artificial intelligence models or reinforcement learning algorithms.

Effect of distance metrics on positioning accuracy / Indoor Positioning Systems

RF fingerprinting based positioning systems have come to the forefront among indoor positioning systems as they provide both high precision and can use existing infrastructure like Wi-Fi access points effectively. In this study, an architecture that constructs the RF signal map of the environment in real-time is proposed. Therefore, with the positioning request, both the fingerprint of the target terminal and RF signal map of the environment are obtained in an online manner. At the stage of comparing the RF fingerprint of the target terminal with RF signal map of the environment, effects of different distance metrics (Manhattan, Euclidean, Chebyshev, Canberra, and Sorensen) and location estimation methods on the positioning accuracy were inspected. It was shown through experiments that the Chebyshev is the best distance metric among others. Moreover, it was seen that location estimation methods have no major impact on positioning accuracy.

2018 26th Signal Processing and Communications Applications Conference (SIU)

Evolutionary Reinforcement Learning for the Coordination of Swarm UAVs

Deep Reinforcement Learning (DRL) algorithms are used in many challenging tasks and their usage areas are rapidly increasing. One of these areas is the formation flights of Unmanned Aerial Vehicles (UAVs). The rising of Reinforcement Learning (RL) algorithms performances is directly proportional to the development of environments. This paper presents a new environment developed through software (Ardupilot, Mavlink, drone-kit) that is frequently used in open source UAV simulation and programming, and the performance of the Evolutionary Reinforcement Learning (ERL) agent in this environment. The difference of this environment is that, unlike other environments, the model can be operated directly on a drone-kit supported vehicle and is specifically defined on the centralised formation task. The aim of this study is; in order to question the performance of the Evolutionary Reinforcement Learning (ERL) algorithm which has better results than other algorithms in DRL training environments,in this environment, and increasing the usage of the algorithm in this direction.

2020 28th Signal Processing and Communications Applications Conference (SIU)

Deep Learning Inference Acceleration Library for Tf / Pytorch

While deep learning applications are increasing rapidly, the acceleration of these applications on Field-Programmable Gate Arrays (FPGAs) has become very popular. Deep learning applications demos are usually made using high-level Python libraries such as Tensorflow and Pytorch.The transfer of these deep learning models written in high level languages to FPGAs requires expertise and long work times.In this article, we present an open source tool that maps deep learning models written in Tensorflow and Pytorch to System-on-Chip (SoC) solution.

On Going