Seminar on selected topics in advanced networks and applications

Thursday, December 03, 2020, Zoom 

Higher School of Economics (HSE), Moscow Institute of Electronics and Mathematics, Alexey Kreschuk: On the Capacity Estimation of a Slotted Multiuser Communication Channel

Abstract
We consider a vector-disjunctive channel where users transmit some vector of bits of length L. We estimate the capacity of this channel and derive the lower bound on this value. And then present some numerical results for the bound we obtained in the case of different number of active users and different parameters of channel.

The speaker, Alexey Kreschuk has defended his PhD Thesis in Institute for Information Transmission Problems, Russian Academy of Science. He works as Associate Professor and Senior Researcher in the Higher School of Economics since 2017 continuing active research in Algebraic and Combinatorial Coding Theory, LDPC Coding, Multiple access systems. Dr. Kreschuk is also participating in HSE-Huawei joint R&D projects focusing on development of new effective PHY for fiber optic communication systems

Brno University of Technology, Czechia, Pavel Seda: AI-aided Anomaly Detection on Smart Home Data Sets

Abstract
Nowadays, it is quite common for ordinary households to have installed a vast amount of IoT and/or Smart Home devices. Using these devices, we are able to manage our households without any difficulties and enjoy the benefits resulting from the fully connected ecosystems. However, we must not forget that in addition to the control of the household itself, we can integrate artificial intelligence methods that can detect unexpected behavior and thus help us prevent unfortunate events or otherwise effectively manage our household. Through this talk, we are going to discuss the results of our work consisting of the design and integration of anomaly detection service for a Smart Home prototype created within a project between the Brno University of Technology and A1 Telekom Austria Group company.

Pavel Seda received his MSc. degree in electrical engineering from the Faculty of Electrical Engineering and Communication at the Brno University of Technology (BUT), the Czech Republic in 2017 and also MSc. in Applied Informatics from Masaryk University in 2018. From 2014 to 2018 Pavel worked as Java Developer at the IBM Czech Republic, in one of its developer divisions. Currently, Pavel is acting as the PhD candidate at BUT and his research topics include different areas of Artificial intelligence especially with the focus on heuristic algorithms on location covering problems.

University of Coimbra, Portugal, Marco Silva: Heuristic and Simulation of Energy Harvesting IoT Networks

Abstract
The advances made in sustainable energy technology to power small Internet-of-Things (IoT) devices have raised the challenge of how to prolong the Energy Neutral Operation (ENO), in which no node depletes its battery and all the nodes prevent batteries from overcharging. Extending the operational time of IoT networks through environmentally-friendly sources (e.g. solar, thermal, and vibration) is the first stage towards sustainable energy communications, but it is still a challenge to harvest energy from sustainable sources and distributively control the consumption to avoid over or under usage. Therefore, optimization and distributed heuristics should be proposed to achieve and maintain ENO as long as possible. This work sets out an optimization model and a distributed heuristic that is able to achieve and prolong ENO in IoT networks, by supporting periodic and event-based traffic data. In addition to this, the proposed solutions achieve ENO by controlling the resulting traffic and transmissions, including the use of data aggregation. The simulation that was conducted shows the benefits of the distributed heuristics study in terms of time in ENO and amount of event-traffic delivered to the sink node.

Marco Silva, holds a Master degree in Electrical and Computer Engineering, with a specialization in Computers at the University of Coimbra. He is currently a PhD student at the University of Coimbra. Driven by curiosity, after his Bachelor's degree, he joined a research project at the Centre for Informatics and Systems of the University of Coimbra (CISUC) in the field of intelligent sensing for 5G platforms, having been a researcher on the project P2020 MobiWise. His research interests involve intelligent control mechanisms in communication networks, energy efficiency and 5G networks.

Tampere University, Finland, Roman Kovalchukov: Infinitely-scalable decentralized routing for wireless mesh networks

Abstract
The wireless mesh network is a key solution for multiple future technologies, including smartphone ad hoc networks, flying ad hoc networks, and disaster rescue ad hoc networks. Our research project aims to develop theoretical foundations and propose practical algorithms for topology self-organization and maintenance of scalable data routing solutions for communication and control in dynamic wireless mesh networks. We will solve this problem by developing a new type of Virtual Coordinate System (VCS). Our VCS mimics wireless network nodes' physical positions without knowing their true coordinates, based on connectivity and wireless channel characteristics. Utilizing the VCS, we can route the information from source to destination based on their virtual coordinates by propagating messages towards the target node. 

Roman Kovalchukov received an M.S. degree in Information Technology at Tampere University (TAU), Finland, in 2020. Currently, he is a Doctoral Researcher at TAU, pursuing a Ph.D. degree. His research interests are in wireless communications, focusing on the analysis of mmWave cellular networks, wireless mMTC, and mesh networks using stochastic geometry, graph theory, machine learning, queueing theory, and system-level simulations.    

Higher School of Economics (HSE), Moscow Institute of Electronics and Mathematics, Anton Sergeev: Human detection on thermal images: comparison of SSD, RCNN and TSMO-based algorithms

Abstract
The goal of the R&D project is to develop efficient algorithms and system for detection of people (vertical human figures) in the thermal images (from thermal infrared camera) in real time. The results of the work could be used in pedestrian detection, traffic management systems, smart- and self-driving cars, safety and security solutions  and etc. The research is focused on industrial usage cases. The resultant system should be installed on large truck for transporting heavy loads, which are used at the remote areas of Siberia and Middle Asia at oil producing and processing facilities, container yards, steel-making plants. Most of this facilities are located in the remote northern areas with long-long dark winters and bad rainy/snowy weather conditions. This leads to "reduced visibility" for drivers, killing and injuring people on the road which are not detected on time.

The first phase of the research is to compare the efficiency (sensitivity, specificity) of different computer vision and ML approaches in the area of human detection on thermal images. The presentation shows the experimental results of this comparison for TSMO-based, Triple-histogram, SSD, RCCN algorithms for 2 totally different data sets (1500+2000 images).

The research supervisor, Anton Sergeev has joint HSE in 2017. As advisor, he is responsible for managing and organizing R&D projects with HSE national and international partners (Huawei, Infowatch, Infotecs, Rosatom, WorldSkills Russia/International, Megaphone etc.). Before HSE he had worked as CIO in St. Petersburg State University of Aerospace Instrumentation, responsible for providing all aspects of IT systems in the University (including support, development, business improvement and systems’ evolution) and research services for external clients (Intel, Nokia, NSN, Siemens, DellEMC etc.). His research interests include information security, data leakage protection, wireless communications, machine learning.

Higher School of Economics (HSE), Moscow Institute of Electronics and Mathematics, Rostislav Shaniiazov: Research of coding-modulation scheme considering statistical properties of a data source

Abstract
This research concerns new effective method of joint data coding/modulation using constellation shaping. It may improve energy-efficiency and energy savings of modern wireless transmission systems. The method require a priori knowledge of input data probability distribution to map them to the modulation symbols in the most efficient way. We present the comparison of capacity and theoretical and practical bit error rate with shaping and a conventional information transmission scheme.

Rostislav Shaniiazov is a PhD student of HSE. He graduated MSc Skolkovo Institute of Science and technology and Saints Petersburg Aerospace Instrumentation in 2019.