Loukas Petrou

A Bayesian Multiple Models Combination Method for Time Series Prediction

In this paper we present the Bayesian Combined Predictor (BCP), a probabilistically motivated predictor for time series prediction. BCP utilizes local predictors of several types (e.g., linear predictors, artificial neural network predictors, polynomial predictors etc.) and produces a final prediction which is a weighted combination of the local predictions; the weights can be interpreted as Bayesian posterior probabilities and are computed online. Two examples of the method are given, based on real world data: (a) short term load forecasting for the Greek Public Power Corporation dispatching center of the island of Crete, and (b) prediction of sugar beet yield based on data collected from the Greek Sugar Industry. In both cases, the BCP outperforms conventional predictors.

A robotic system for handling textile and non rigid flat materials

Abstract This paper presents a robotic system incorporating vision and force/torque sensing for handling flat textile materials. The methodology followed consists in tackling representative problems related to a broad range of materials. Representative tasks have been selected which are further analyzed into simple operations. Algorithms for sensory data processing and control have been developed. Finally, conclusions from experimentation are presented.

Robot handling of flat textile materials

The authors present a robotic system incorporating vision and force/torque sensing for handling flat textile materials and describe the results of experiments to measure the accuracy and reliability of the system for a variety of representative handling tasks for textile materials.

A robotic system for handling textile materials

This paper presents a robot system incorporating vision and force/torque sensing for handling of flat textile materials. Experimentation is used to draw conclusions concerning the performance of standard arms and sensing techniques. Also requirements of such systems are discussed.

Critical Rays Scan Match SLAM

Scan matching is one of the oldest and simplest methods for occupancy grid based SLAM. The general idea is to find the pose of a robot and update its map simply by calculating the 2-D transformation between a laser scan and its predecessor. Due to its simplicity many solutions were proposed and used in various systems, the vast majority of which are iterative. The fact is, that although scan matching is simple in its implementation, it suffers from accumulative noise. Of course, there is certainly a trade-off between the quality of results and the execution time required. Many algorithms have been introduced, in order to achieve good quality maps in a small iteration time, so that on-line execution would be achievable. The proposed SLAM scheme performs scan matching by implementing a ray-selection method. The main idea is to reduce complexity and time needed for matching by pre-processing the scan and selecting rays that are critical for the matching process. In this paper, several different methods of ray-selection are compared. In addition matching is performed between the current scan and the global robot map, in order to minimize the accumulated errors. RRHC (Random Restart Hill Climbing) is employed for matching the scan to the map, which is a local search optimization procedure that can be easily parameterized and is much faster than a traditional genetic algorithm (GA), largely because of the low complexity of the problem. The general idea is to construct a parameterizable SLAM that can be used in an on-line system that requires low computational cost. The proposed algorithm assumes a structured civil environment, is oriented for use in the RoboCup - RoboRescue competition, and its main purpose is to construct high quality maps.

An FPGA implementation of the SMG-SLAM algorithm

One of the main tasks of a mobile robot in an unknown environment is to build and update a map of the environment and simultaneously determine its location within this map. This problem is referred to as the simultaneous localization and mapping (SLAM) problem. The article introduces scan-matching genetic SLAM (SMG-SLAM), a novel SLAM algorithm. It is based on a genetic algorithm that uses scan-matching for gene fitness evaluation. The main scope of the article is to present a hardware implementation of SMG-SLAM using an field programmable gate array (FPGA). The architecture of the system is described and it is shown that it is up to 14.83 times faster compared to the software algorithm without significant loss in accuracy. The proposed implementation can be used as part of a larger system, providing efficient SLAM for autonomous robotic applications.

Multi-objective optimization for dynamic task allocation in a multi-robot system

In this paper, we propose a solution to the Multi-Robot Dynamic Task Allocation problem. We use Multi-Objective optimization in order to estimate, and subsequently, make an offer for its assignment. The motivation is to provide a generic solution, independent of the domain, with an aim to better utilize resources such as time or energy. The algorithm provides a significant degree of flexibility, and can be implemented in a number of diverse domains, provided the modeling of the parameters follows the convention presented. For this, we take into account - besides the distance traveled - the efficiency of a robot in a specific task type. The system has been shown to demonstrate scalability, as the experimental results indicate. It is also capable of responding to changes in the environment.

Special-purpose architectures for fuzzy logic controllers

Abstract The design of two dedicated architectures for real-time fuzzy inference are proposed and discussed. The first is designed for versatility, being able to handle fuzzy logic controller implementations with arbitrary size parameters. The second is a highly parallel structure, based on a scheme for the reduction of complexity of the fuzzy inference algorithm. The bottleneck of the defuzzification process is overcome by a proposed method for parallel calculation.

A Review of Global Path Planning Methods for Occupancy Grid Maps Regardless of Obstacle Density

Path planning constitutes one of the most crucial abilities an autonomous robot should possess, apart from Simultaneous Localization and Mapping algorithms (SLAM) and navigation modules. Path planning is the capability to construct safe and collision free paths from a point of interest to another. Many different approaches exist, which are tightly dependent on the map representation method (metric or feature-based). In this work four path planning algorithmic families are described, that can be applied on metric Occupancy Grid Maps (OGMs): Probabilistic RoadMaps (PRMs), Visibility Graphs (VGs), Rapidly exploring Random Trees (RRTs) and Space Skeletonization. The contribution of this work includes the definition of metrics for path planning benchmarks, actual benchmarks of the most common global path planning algorithms and an educated algorithm parameterization based on a global obstacle density coefficient.

A Computational Model for Path Loss in Wireless Sensor Networks in Orchard Environments

A computational model for radio wave propagation through tree orchards is presented. Trees are modeled as collections of branches, geometrically approximated by cylinders, whose dimensions are determined on the basis of measurements in a cherry orchard. Tree canopies are modeled as dielectric spheres of appropriate size. A single row of trees was modeled by creating copies of a representative tree model positioned on top of a rectangular, lossy dielectric slab that simulated the ground. The complete scattering model, including soil and trees, enhanced by periodicity conditions corresponding to the array, was characterized via a commercial computational software tool for simulating the wave propagation by means of the Finite Element Method. The attenuation of the simulated signal was compared to measurements taken in the cherry orchard, using two ZigBee receiver-transmitter modules. Near the top of the tree canopies (at 3 m), the predicted attenuation was close to the measured one—just slightly underestimated. However, at 1.5 m the solver underestimated the measured attenuation significantly, especially when leaves were present and, as distances grew longer. This suggests that the effects of scattering from neighboring tree rows need to be incorporated into the model. However, complex geometries result in ill conditioned linear systems that affect the solvers convergence.