Intelligent Systems and Networking: research areas

Optimisation and Decision Support

Modeling, simulation and optimization are fundamental to solve problems in a number of fields of science, technology and life.
The main goal is to design, implement, simulate and optimize dynamic and complex systems. Simulation, the exploration of the dynamic behavior of the model in time and space, is investigated for both continuous and discrete-event systems. Simulating a model allows the evaluation of indicators of the performance of the modeled system, improving our understanding of its behavior and dynamic complexity.
A number of optimization techniques are approached and thought to explore such performance space in order to find the best way of managing the system under study. In case of large search spaces, other than exploring the entire search space the goal is to smart select only a subset of possible solutions using intelligent and adaptive heuristics able to learn from experience and to produce good results in short time. These modern heuristics (such as Simulated Annealing; Ant Colony Optimisation, Tabu Search, Genetic Algorithms; GRASP; Particle Swarm Optimisation) combines principles coming from biological systems, artificial intelligence, computer science and operational research.

Machine Learning and Data Mining

Machine learning arise in a wide range of modern applications: web search, robotics & embedded computing, pattern recognition, scientific computing, scheduling, optimization, etc.
Foundations of machine learning and data mining as well as asymptotically optimal general problem solvers are investigated. In the research and teaching activities fundamentals topics related to machine learning systems and data mining are covered: Bayesian / probabilistic reasoning; hidden Markov models; expectation maximization; neural networks; support vector machines; reinforcement learning; unsupervised learning techniques, data mining, pattern classification and regression, empirical evaluations, feature selection, discretization, combining multiple models, cluster analysis.

Intelligent Networking (opportunistic, mesh and ad hoc networking)

Multi-hop ad hoc networking is an area of computer networking that encountered large interest from both civilian and industrial research.
Some classes of multi-hop ad hoc networks, as sensor networks and vehicular networks, have already found their place in the market and are widely used. They are however evolving in terms of applications and technologies. Other classes are starting to enter the market, as mesh networks and opportunistic networks, demonstrating their validity as flexible and “low cost” extension of the Internet.
Students graduating from this program face the evolution of networking and work with academic and industrial aspects of designing and developing such networks in different computing environment. An interdisciplinary approach, which involves both the application and the technology aspects, allows to develop and apply this knowledge to real scenarios that correspond to the market needs to date and in the future.
The work performed in this area includes theoretical work on routing, mobility, resource management and is complemented by a practical part (simulation, experiment, and trial) in the networking lab. Experiments and trials are performed with novel devices and platforms as sensor networks, smart phones, wi-fi and mesh routers.  Interaction with major industries in the telecommunication area, within the framework of the international projects of this area, is also foreseen.

Robotics

Understanding the foundation of robotics is essential for building complete knowledge of artificial systems. Intelligence appears when a physical life-form is interacting with the environment and it can only be observed in an interactive process. Artificial systems cannot be fully understood by simulating these systems, because generalization from real embodies systems to simulated systems leads always to an oversimplification of the real world and its challenges. Upon completion, students learn about the different classes of robots, their application areas and their varying concepts of programming and problem solving strategies. Topics: Overview of Robotics in Practice and Research; Sensors and Actuators; Low Level Feedback Control; Computer Vision for Robotics; Robotic Manipulation; Legged Locomotion; High Level Behavior Control; Planning, Execution and Learning; Navigation, Localization and Mapping; Sensor Fusion, Communication and Swarm Robotics.

The learned theoretical skills is applied to projects in the robotic laboratory, helpful for the full understanding of the concepts and for inspiring students’ own ideas and further research. The lab is equipped with many robot prototype, 20 E-TUCK Robot, 4 Pioneer3 AT Robot, 2 Neuronics Robot CATANA 6M and 1 Robertino Robot.

Semantic Information Processing

Semantic information processing focuses on the integration of different fields mainly including ontology engineering, ontological text mining, and Semantics Web technologies. Its main applications are the generation and use of semantic information repositories aiming to improving decision making. In current ICT systems, information retrieval weakness is mainly due to the absence of context or semantics in Web information, which makes the current Web oriented to humans and not to computers.
The main innovation of the Semantic Web is the provision of notations or languages to express content semantics, including RDF, OWL, Topic Maps, etc., in the form of metadata and inference rules. These languages may be used to classify and catalogue contents so that they could be automatically organized and filtered by agents, such as with RSS channels, items and feeders. The languages for content formalization and representation are nowadays stable and well defined W3C Specifications in widespread use. In this way the resources (information, knowledge, services, etc.) are explicitly or implicitly stored in a multi-dimensional semantic web based world of relationships and concepts.
While the process of content formalization in terms of Web resources can be accomplished by using these languages, another issue concerns how to manage and present this information. There are two complementary approaches: the first one is to turn machines in new types of Web users able to understand the meaning of data and what to do with them; the second is to make the Web more useful for human beings by presenting information in a comprehensible visual form. Both processing and visualizing semantic information will be determinant for the emergence of an Intelligent Web understandable to human and machines. The Intelligent Web can be considered as the integration of Web 2.0 and Web 3.0, where the rich interaction paradigms of the former are exploited in order to provide a presentation and user interface layer to the latter.

Master Research Unit