Research Interests:
Research Interests:
Research Interests:
Research Interests:
Research Interests:
Research Interests:
Research Interests:
First, Das and colleagues considered only variables with ap value ≤0·05 in univariate analysis for inclusion in the MLR models. However, it is generally recommended that all variables with ap value of 0·25 in univariate analysis, or those... more
First, Das and colleagues considered only variables with ap value ≤0·05 in univariate analysis for inclusion in the MLR models. However, it is generally recommended that all variables with ap value of 0·25 in univariate analysis, or those regarded as being clinically important, ...
Research Interests:
Research Interests:
Research Interests:
The United Nations Sustainable Development Goal 12 is the foundation of the European University on Responsible Consumption and Production consortium (EURECA-PRO). This article introduces the eight EURECA-PRO partner universities:... more
The United Nations Sustainable Development Goal 12 is the foundation of the European University on Responsible Consumption and Production consortium (EURECA-PRO). This article introduces the eight EURECA-PRO partner universities: Montanuniversität Leoben (Austria), Technische Universität Bergakademie Freiberg (Germany), Technical University of Crete (Greece), University of León (Spain), Silesian University of Technology (Poland), Mittweida University of Applied Sciences (Germany), University of Petroşani (Romania), and Hasselt University (Belgium). In addition, each university’s role within the alliance and unique research and study programmes are outlined. The synergy created by EURECA-PRO enables the pursuit of an ambitious research agenda with five research “Lighthouse Missions” as well as the implementation of joint study programmes.
Research Interests:
maida©cacs, usl. edu Neural maps have been recently proposed as an alter-native method for mobile robot path planning (Glasius, Komoda, and Gielen 1995). However, these proposals are mostly theoretical and are primarily concerned with... more
maida©cacs, usl. edu Neural maps have been recently proposed as an alter-native method for mobile robot path planning (Glasius, Komoda, and Gielen 1995). However, these proposals are mostly theoretical and are primarily concerned with biological plausibility. Our purpose is to investigate their applicability on real robots. Information about the environment is mapped on a topologically ordered neural population. The diffusion dynamics force the network into a unique equilibrium state that defines the navigation landscape for the given target. A path from any initial position to the target (corresponding to the peak of the activation surface) is derived by a steepest ascent procedure. The figures below show an example on a 50 x 50 rectangular map (a. Environment, b. Contours of activation, c. Path). We attempted to implement the approach on a No-mad 200 mobile robot for sonar-based navigation. How-ever, we found that the neural map requires reorgani-zation in a polar topology that re...
decision support system to facilitate management of patients with acute gastrointestinal bleeding.
Planning has been one of the main research areas in AI. For about three decades AI researchers explore alternative paths to build intelligent agents with advanced planning capabilities. However, the classical AI planning techniques suffer... more
Planning has been one of the main research areas in AI. For about three decades AI researchers explore alternative paths to build intelligent agents with advanced planning capabilities. However, the classical AI planning techniques suffer from inapplicability to real world domains, due to several assumptions adopted to facilitate research. Attempts to apply planning into real domains must address the problem of uncertainty, which requires a revision of the classical planning framework. Probabilistic models seem to offer a promising alternative, providing models of planning where plans can be represented, generated and evaluated under a standard probabilistic interpretation of uncertainty. This survey paper 1 attempts to cover the recent work in this direction and trigger the interest of the reader for further study and exploration.
Research Interests:
Planning has been one of the main research areas in AI. For about three decades AI researchers explore alternative paths to build intelligent agents with advanced planning capabilities. However, the classical AI planning techniques suffer... more
Planning has been one of the main research areas in AI. For about three decades AI researchers explore alternative paths to build intelligent agents with advanced planning capabilities. However, the classical AI planning techniques suffer from inapplicability to real world domains, due to several assumptions adopted to facilitate research. Attempts to apply planning into real domains must address the problem of uncertainty, which requires a revision of the classical planning framework. Probabilistic models seem to offer a promising alternative, providing models of planning where plans can be represented, generated and evaluated under a standard probabilistic interpretation of uncertainty. This survey paper 1 attempts to cover the recent work in this direction and trigger the interest of the reader for further study and exploration.
Research Interests:
Research Interests: Engineering and League
First, Das and colleagues considered only variables with ap value ≤0·05 in univariate analysis for inclusion in the MLR models. However, it is generally recommended that all variables with ap value of 0·25 in univariate analysis, or those... more
First, Das and colleagues considered only variables with ap value ≤0·05 in univariate analysis for inclusion in the MLR models. However, it is generally recommended that all variables with ap value of 0·25 in univariate analysis, or those regarded as being clinically important, ...
Research Interests:
The Spare Allocation problem (or equivalently Vertex Cover in bipartite graphs) deals with the optimal allocation of spare rows and columns over a two-dimensional array of cells, some of which are faulty. The purpose is to repair all the... more
The Spare Allocation problem (or equivalently Vertex Cover in bipartite graphs) deals with the optimal allocation of spare rows and columns over a two-dimensional array of cells, some of which are faulty. The purpose is to repair all the faulty cells using spares with the minimum possible overall cost. In order to solve the problem optimally, a branch and bound algorithm is employed. Several heuristics for pruning the search tree and a heuristic that estimates the final cost from partial solutions are used to inform and guide the search procedure, leading to an effective A* search algorithm. However, memory limitations on one hand and the good enough quality of the heuristic function that low bounds the cost, on the other, force an Iterative Deepening implementation of the algorithm (IDA * ). Moving in the parallel implementation, many complications arise. A * is by its nature sequential and attempts to parallelize it, led to an amount of redundant search and/or early memory overflows. IDA * provides for parallelism, but work balancing issues should be addressed carefully. The final parallel IDA * search algorithm is scalable and leads to a significant amount of speedup, as demonstrated by the experimental results.
Research Interests:
Academia has entered a new teaching, learning, and researching era: an era in which more and more services turn to digital and online forms, distances are eliminated, geographical borders disappear, and telepresence becomes common. Though... more
Academia has entered a new teaching, learning, and researching era: an era in which more and more services turn to digital and online forms, distances are eliminated, geographical borders disappear, and telepresence becomes common. Though accelerated by the pandemic of the last two years, this transition has been in progress for some time. The importance of creatively nurturing students, academic, and scientific staff in the realms of education, practical knowledge, skills, and competence growth has only increased. Investing in best practices in this digital world, both in teaching and in research, supports a connection between the academic world and society at large, raises societal, environmental awareness, and promotes innovation and excellence at all levels. Each of these considerations plays an important role for the EURECA-PRO European University Alliance, a group of eight partner universities from different European countries working together to establish a modern, diverse Eu...
Research Interests: Excellence and Alliance
Uncrewed aerial vehicles (UAVs) are continuously gaining popularity in a wide spectrum of applications, while their positioning and navigation most often relies on Global Navigation Satellite Systems (GNSS). However, numerous conditions... more
Uncrewed aerial vehicles (UAVs) are continuously gaining popularity in a wide spectrum of applications, while their positioning and navigation most often relies on Global Navigation Satellite Systems (GNSS). However, numerous conditions and practices require UAV operation in GNSS-denied environments, including confined spaces, urban canyons, vegetated areas and indoor places. For the purposes of this study, an integrated UAV navigation system was designed and implemented which utilizes GNSS, visual, depth and inertial data to provide real-time localization. The implementation is built as a package for the Robotic Operation System (ROS) environment to allow ease of integration in various systems. The system can be autonomously adjusted to the flight environment, providing spatial awareness to the aircraft. This system expands the functionality of UAVs, as it enables navigation even in GNSS-denied environments. This integrated positional system provides the means to support fully auto...
Research Interests:
Humanitarian Crisis scenarios typically require immediate rescue intervention. In many cases, the conditions at a scene may be prohibitive for human rescuers to provide instant aid, because of hazardous, unexpected, and human threatening... more
Humanitarian Crisis scenarios typically require immediate rescue intervention. In many cases, the conditions at a scene may be prohibitive for human rescuers to provide instant aid, because of hazardous, unexpected, and human threatening situations. These scenarios are ideal for autonomous mobile robot systems to assist in searching and even rescuing individuals. In this study, we present a synchronous ground-aerial robot collaboration approach, under which an Unmanned Aerial Vehicle (UAV) and a humanoid robot solve a Search and Rescue scenario locally, without the aid of a commonly used Global Navigation Satellite System (GNSS). Specifically, the UAV uses a combination of Simultaneous Localization and Mapping and OctoMap approaches to extract a 2.5D occupancy grid map of the unknown area in relation to the humanoid robot. The humanoid robot receives a goal position in the created map and executes a path planning algorithm in order to estimate the FootStep navigation trajectory for ...
Research Interests:
Research Interests: Engineering, Computer Science, Artificial Intelligence, Technology, Reinforcement Learning, and 13 moreApproximation Algorithms, Training, Optimization, Denoising, Noise reduction, Degradation, Encoding, Scalability, Science Technology, Dialogue Systems, Electrical And Electronic Engineering, Telecommunications and Networking, and Variational Autoencoders
Research Interests:
A global maximum power point tracking (GMPPT) process must be applied for detecting the position of the GMPP operating point in the minimum possible search time in order to maximize the energy production of a photovoltaic (PV) system when... more
A global maximum power point tracking (GMPPT) process must be applied for detecting the position of the GMPP operating point in the minimum possible search time in order to maximize the energy production of a photovoltaic (PV) system when its PV array operates under partial shading conditions. This paper presents a novel GMPPT method which is based on the application of a machine-learning algorithm. Compared to the existing GMPPT techniques, the proposed method has the advantage that it does not require knowledge of the operational characteristics of the PV modules comprising the PV system, or the PV array structure. Additionally, due to its inherent learning capability, it is capable of detecting the GMPP in significantly fewer search steps and, therefore, it is suitable for employment in PV applications, where the shading pattern may change quickly (e.g., wearable PV systems, building-integrated PV systems etc.). The numerical results presented in the paper demonstrate that the ti...
Research Interests:
This paper presents results on a user interface model for providing universal access to mobile computing devices.
We are interested in finding near-optimal solutions to hard optimization problems using Hopfield-type neu- ral networks. The methodology is based on a basic property of such networks, that of reducing their 'en- ergy' during... more
We are interested in finding near-optimal solutions to hard optimization problems using Hopfield-type neu- ral networks. The methodology is based on a basic property of such networks, that of reducing their 'en- ergy' during evolution, leading to a local or global minimum. The methodology is presented and several different network models usually employed as optimiz- ers (Analog Hopfield net with
Research Interests:
Research Interests:
Research Interests:
Research Interests:
Research Interests:
Research Interests:
Research Interests:
Research Interests:
Research Interests:
Page 157. DIM ACS Series in Discrete Mathematics and Theoretical Computer Science Volume 54. z000 2D DNA Self-Assembly for Satisfiability Michail G. Lagoudakis and Thomas H. LaBean ABSTRACT. DNA self-assembly ...
Research Interests:
Research Interests:
In the field of sequential decision making and reinforcement learning, it has been observed that good policies for most problems exhibit a significant amount of structure. In practice, this implies that when a learning agent discovers an... more
In the field of sequential decision making and reinforcement learning, it has been observed that good policies for most problems exhibit a significant amount of structure. In practice, this implies that when a learning agent discovers an action is better than any other in a given state, this action actually happens to also dominate in a certain neighbourhood around that state. This paper presents new results proving that this notion of locality in action domination can be linked to the smoothness of the environment’s underlying stochastic model. Namely, we link the Lipschitz continuity of a Markov Decision Process to the Lispchitz continuity of its policies’ value functions and introduce the key concept of influence radius to describe the neighbourhood of states where the dominating action is guaranteed to be constant. These ideas are directly exploited into the proposed Localized Policy Iteration (LPI) algorithm, which is an active learning version of Rollout-based Policy Iteration. Preliminary results on the Inverted Pendulum domain demonstrate the viability and the potential of the proposed approach.
Research Interests:
Research Interests:
Research Interests:
ABSTRACT Several recent learning approaches in decision making under uncertainty suggest the use of classifiers for representing policies compactly. The space of possible policies, even under such structured representations, is huge and... more
ABSTRACT Several recent learning approaches in decision making under uncertainty suggest the use of classifiers for representing policies compactly. The space of possible policies, even under such structured representations, is huge and must be searched carefully to avoid computationally expensive policy simulations (rollouts). In our recent work, we proposed a method for directed exploration of policy space using support vector classifiers, whereby rollouts are directed to states around the boundaries between different action choices indicated by the separating hyperplanes in the represented policies. While effective, this method suffers from the growing number of support vectors in the underlying classifiers as the number of training examples increases. In this paper, we propose an alternative method for directed policy search based on relevance vector machines. Relevance vector machines are used both for classification (to represent a policy) and regression (to approximate the corresponding relative action advantage function). Classification is enhanced by anomaly detection for accurate policy representation. Exploiting the internal structure of the regressor, we guide the probing of the state space only to critical areas corresponding to changes of action dominance in the underlying policy. This directed focus on critical parts of the state space iteratively leads to refinement and improvement of the underlying policy and delivers excellent control policies in only a few iterations, while the small number of relevance vectors yields significant computational time savings. We demonstrate the proposed approach and compare it with our previous method on standard reinforcement learning domains (inverted pendulum and mountain car).
Research Interests:
Research Interests: RoboCup and Robot Soccer
ABSTRACT The RoboCup competition is the international robotic soccer world cup organized annually since 1997. The initial conception by Hiroaki Kitano in 1993 led to the formation of the RoboCup Federation with a bold vision: By the year... more
ABSTRACT The RoboCup competition is the international robotic soccer world cup organized annually since 1997. The initial conception by Hiroaki Kitano in 1993 led to the formation of the RoboCup Federation with a bold vision: By the year 2050, to develop a team of fully autonomous humanoid robots that can win against the human world soccer champions! RoboCup poses a real-world challenge for Artificial Intelligence, which requires addressing simultaneously the core problems of perception, cognition, action, and coordination under real-time constraints. In this talk, I will outline the vision, the challenges, and the contribution of the RoboCup competition in its short history. I will also offer an overview of the research efforts of team Kouretes, the RoboCup team of the Technical University of Crete, on topics ranging from complex motion design, efficient visual recognition, and self-localization to robotic software engineering, distributed communication, skill learning, and coordinated game play. My motivation is to inspire researchers and students to form teams with the goal of participating in the various leagues of this exciting and challenging benchmark competition and ultimately contributing to the advancement of the state-of-the-art in Artificial Intelligence and Robotics.
ABSTRACT The ability of learning is critical for agents designed to compete in a variety of two-player, turn-taking, tactical adversarial games, such as Backgammon, Othello/Reversi, Chess, Hex, etc. The mainstream approach to learning in... more
ABSTRACT The ability of learning is critical for agents designed to compete in a variety of two-player, turn-taking, tactical adversarial games, such as Backgammon, Othello/Reversi, Chess, Hex, etc. The mainstream approach to learning in such games consists of updating some state evaluation function usually in a Temporal Difference (TD) sense either under the MiniMax optimality criterion or under optimization against a specific opponent. However, this approach is limited by several factors: (a) updates to the evaluation function are incremental, (b) stored samples from past games cannot be utilized, and (c) the quality of each update depends on the current evaluation function due to bootstrapping. In this paper, we present a learning approach based on the Least-Squares Policy Iteration (LSPI) algorithm that overcomes these limitations by focusing on learning a state-action evaluation function. The key advantage of the proposed approach is that the agent can make batch updates to the evaluation function with any collection of samples, can utilize samples from past games, and can make updates that do not depend on the current evaluation function since there is no bootstrapping. We demonstrate the efficiency of the LSPI agent over the TD agent in the classical board game of Othello/Reversi.
ABSTRACT Several recent learning approaches based on approximate policy iteration suggest the use of classifiers for representing policies compactly. The space of possible policies, even under such structured representations, is huge and... more
ABSTRACT Several recent learning approaches based on approximate policy iteration suggest the use of classifiers for representing policies compactly. The space of possible policies, even under such structured representations, is huge and must be searched carefully to avoid computationally expensive policy simulations (rollouts). In our recent work, we proposed a method for directed exploration of policy space using support vector classifiers, whereby rollouts are directed to states around the boundaries between different action choices indicated by the separating hyper planes in the represented policies. While effective, this method suffers from the growing number of support vectors in the underlying classifiers as the number of training examples increases. In this paper, we propose an alternative method for directed policy search based on relevance vector machines. Relevance vector machines are used both for classification (to represent a policy) and regression (to approximate the corresponding relative action advantage function). Exploiting the internal structure of the regress or, we guide the probing of the state space only to critical areas corresponding to changes of action dominance in the underlying policy. This directed focus on critical parts of the state space iteratively leads to refinement and improvement of the underlying policy and delivers excellent control policies in only a few iterations, while the small number of relevance vectors yields significant computational time savings. We demonstrate the proposed approach and compare it with our previous method on standard reinforcement learning domains.
Research Interests:
ABSTRACT The design of complex dynamic motions for humanoid robots is achievable only through the use of robot kinematics. In this paper, we study the problems of forward and inverse kinematics for the Aldebaran NAO humanoid robot and... more
ABSTRACT The design of complex dynamic motions for humanoid robots is achievable only through the use of robot kinematics. In this paper, we study the problems of forward and inverse kinematics for the Aldebaran NAO humanoid robot and present a complete, exact, analytical solution to both problems, including a software library implementation for realtime onboard execution. The forward kinematics allow NAO developers to map any configuration of the robot from its own joint space to the three-dimensional physical space, whereas the inverse kinematics provide closed-form solutions to finding joint configurations that drive the end effectors of the robot to desired target positions in the three-dimensional physical space. The proposed solution was made feasible through a decomposition into five independent problems (head, two arms, two legs), the use of the Denavit-Hartenberg method, and the analytical solution of a non-linear system of equations. The main advantage of the proposed inverse kinematics solution compared to existing approaches is its accuracy, its efficiency, and the elimination of singularities. In addition, we suggest a generic guideline for solving the inverse kinematics problem for other humanoid robots. The implemented, freely-available, NAO kinematics library, which additionally offers center-of-mass calculations, is demonstrated in two motion design tasks: basic center-of-mass balancing and pointing to the ball.
In the field of sequential decision making and reinforcement learning, it has been observed that good policies for most problems exhibit a significant amount of structure. In practice, this implies that when a learning agent discovers an... more
In the field of sequential decision making and reinforcement learning, it has been observed that good policies for most problems exhibit a significant amount of structure. In practice, this implies that when a learning agent discovers an action is better than any other in a given state, this action actually happens to also dominate in a certain neighbourhood around that state. This paper presents new results proving that this notion of locality in action domination can be linked to the smoothness of the environment’s underlying stochastic model. Namely, we link the Lipschitz continuity of a Markov Decision Process to the Lispchitz continuity of its policies’ value functions and introduce the key concept of influence radius to describe the neighbourhood of states where the dominating action is guaranteed to be constant. These ideas are directly exploited into the proposed Localized Policy Iteration (LPI) algorithm, which is an active learning version of Rollout-based Policy Iteration. Preliminary results on the Inverted Pendulum domain demonstrate the viability and the potential of the proposed approach.
In the field of sequential decision making and reinforcement learning, it has been observed that good policies for most problems exhibit a significant amount of structure. In practice, this implies that when a learning agent discovers an... more
In the field of sequential decision making and reinforcement learning, it has been observed that good policies for most problems exhibit a significant amount of structure. In practice, this implies that when a learning agent discovers an action is better than any other in a given state, this action actually happens to also dominate in a certain neighbourhood around that state. This paper presents new results proving that this notion of locality in action domination can be linked to the smoothness of the environment’s underlying stochastic model. Namely, we link the Lipschitz continuity of a Markov Decision Process to the Lispchitz continuity of its policies’ value functions and introduce the key concept of influence radius to describe the neighbourhood of states where the dominating action is guaranteed to be constant. These ideas are directly exploited into the proposed Localized Policy Iteration (LPI) algorithm, which is an active learning version of Rollout-based Policy Iteration. Preliminary results on the Inverted Pendulum domain demonstrate the viability and the potential of the proposed approach.