Capabilities

We design and and implement algorithms to distinguish known or unknown anomalous behaviors from the normal behavior of the target dynamic system. Our modeling techniques directly process sensor data and utilize variable order Markov models with configurable complexity parameters to obtain both generalizable and robust to noise models of different behaviors. By using our novel model matching function, we calculate the deviation from the normal behavior and detect if there is an unexpected event happening. Further, we identify anomalies by comparing their models to previously seen ones and label them as critical/non-critical. Finally, we predict future anomalies considering the matching ratio/probability and the observed pattern of behavior changes.

Our main focus area is V2X Communications and Connected Cars. On the other hand, we have experience and knowledge on other types of wireless networks which is building our expertise in this area. 5G, V2X (IEEE 802.11p, IEEE 1609, ITS G5, and other ETSI standards for V2X, 3GPP R14-R16), IEEE 802.11 (WiFi), Bluetooth, WSN/ZigBee and other wireless technologies and their standards are the main areas we work and have experience. Simulations and building simulation models are other skills we have.

• Design and development of algorithms
• All types of Wireless Networks (in all layers)
• V2V and V2I, 5G-V2X Communications
• Platooning, Remote Driving
• MAC Layer Algorithms (Bandwidth efficiency, QoS)
• TCP Algorithms (Reliability, Efficiency, QoS)
• Clustering algorithms (Bandwidth Utilization,
  Information Dissemination, QoS)
• Compliant and conforming the standards

• Modelling and Simulation
• City traffic (urban/highway/ traffic lights)
• V2V – V2I – V2N communication
• LTE integrated simulation
• Compliant and conforming the standards
• Veins, SUMO, OpenStreetMap
• Building test environment & Testing
• via simulation
• via test-bed
• via emulators

While we are working on V2X communications, we consider the real vehicle traffics, therefore we also work on vehicle traffic analysis. We have real vehicle traces collected daily, appx. 100.000 vehicles/day. We are able to analyze these traces depending on our needs. We have experience on that. We are able to model the vehicle traffic and extract spatiotemporal characteristics for building the model and its optimization depending on the needs.  These models can be used to estimate the pollution , energy consumption, etc. as well as model fitting. For improving the reliability of the proposed solution and the new system, such models are essential.

Here are some key capabilities and contributions:

• Real Vehicle Trace Analysis of city-wide (>70000 traces per day) and country-wide (>1M traces per day)
• Extracting some parameters for V2X communication
• Obtain spatiotemporal characteristics
• contributes to many areas in communication and resource planning
• some parameters speed, density etc. are easy to obtain
• We are interested in communication perspective (spatiotemporally)
• Inter-vehicular distance
• Inter-contact duration
• Connected components (of vehicles)
• Available bandwidth vs data rate
• Connectivity and coverage (for network and resource management)
• RSU/eNodeB placement, resource scheduling, content delivery, workload offloading

Wireless networks and network performance measurements are one of our major research interests. In addition to the studies on the protocol stack and optimizations, we are also developing methods and tools to measure network performance and collect the necessary data for online and post-processing analysis. Reducing errors and removing noise in data uncovers real network behaviour on various deployments and scenarios. Accurate measurement plays important role to obtain reliable set of data. Detection and identification of outliers, measurement errors and other anomalous conditions become uncomplicated when the data is reliable and accurate. With this intention, we are also developing graphical network performance analyzer that visualizes network performance metrics that allows monitoring in real-time.

Data and statistical results dynamically updated for real-time visualization along with historical data. Estimation models and filters help user to dynamically smooth out noisy data, estimates next value, and generate alerts in case of anomalies. Methods and indicators on the interface are carefully placed for scientific view. Expediently selected charts and graphs show accuracy in data, trends, anomalies, and losses that enhances user's perception. Network & Analyzer settings allows user and experiment-specific modifications in the measurements and their analysis. Well-known, community accepted and reliable software development patterns and tools satisfy reliability, reusability and maintainability requirements.
- Reliability: implementations and methods in data gathering and measurements conforms the standards, RFCs and regulations.
- Reusability: modularity allows user to adapt analyzer components for visualizing different types of data models.
- Maintainability/sustainability: additional, complementary, or new measurement methods can easily be implemented/integrated.

We work on generating accurate regional roadmaps using geolocation data obtained from vehicles. We first derive necessary features e.g velocity and bearing, by projecting raw location data over gnomonic projection considering the great circle distances. Data analysis for data cleaning is one main tasks for accuracy. Clustering and machine learning methods are applied to extract the information/parameters needed. We then utilize these findings to generate trace predictions for vehicles and to generate road map (road infrastructure). These trace predictions and trace clusters of varying densities are translated into roads of different characteristics when displayed on geospatial projection.

We design and implement algorithms to locate assets and people indoors with high accuracy. Gathered data such as Received Signal Strength Indicator (RSSI), angle of received signal, direction and average velocity of the assets are used to locate the assets using techniques such as machine learning, dead reckoning, trilateration and triangulation.

We analyze the location and trace data for predictions and post-processing. We utilize the historical data to generate heat maps in desired time periods in addition to the analysis of behavior of assets and people. We can observe an individual asset's path, most visited locations etc and determine relations between other assets.

Internet of Things(IoT) is a network in which electrical components and actuators are connected and exchange information with each other. Main components of the IoT is sensor devices&actuators, connectivity, data processing and user interface. We focus on all main components to create reliable, secure and sustainable IoT system.

In MInD-NET laboratory we are working on single-board computers and microcontrollers, along with various sensors to gather, process and transfer data. Data is used for feedback & control, changing the state of a system; create reports, make identification and classification. We combine and correlate different variaty and number of sensor data along with historical results to validate, uncover and provide complete information. We are working on and have experience in wired and wireless connectivity in all layers. With this knowledge, we select proper methods and protocols according to the characteristic of the domain.

IoT enables interdisciplinary work. With sensors and data processing, computer science can be included into many fields in industry. We develop applications with MQTT and REST protocols to ensure reliability and security in data transfer from IoT end-device to industry. We specify methods and systems to assure interconnection with the needs and availability of the domain. In the laboratory it is possible to simulate various scenarios for different use cases.