Andre Bolles

Streaming SPARQL extending SPARQL to process data streams

A lot of work has been done in the area of data stream processing. Most of the previous approaches regard only relational or XML based streams but do not cover semantically richer RDF based stream elements. In our work, we extend SPARQL, the W3C recommendation for an RDF query language, to process RDF data streams. To describe the semantics of our enhancement, we extended the logical SPARQL algebra for stream processing on the foundation of a temporal relational algebra based on multi-sets and provide an algorithm to transform SPARQL queries to the new extended algebra. For each logical algebra operator, we define executable physical counterparts. To show the feasibility of our approach, we implemented it within our ODYSSEUS framework in the context of wind power plant monitoring.

SaLsA Streams: Dynamic Context Models for Autonomous Transport Vehicles Based on Multi-sensor Fusion

Due to the fact that currently operating autonomous vehicles can observe only a limited area with their onboard sensors, safety regulations often dictate a very slow speed. However, as more and more sensors in the environment are available, we can fuse their information and provide extended information as a shared context model to support the autonomous vehicles. In this paper, we consider a scenario with a publicly accessible area that is populated with autonomous transport vehicles, human guided vehicles like trucks or bicycles, and pedestrians. We analyze requirements and challenges for highly dynamic context models in this scenario. Furthermore, we propose a comprehensive system architecture that can cope with these challenges, namely deterministic processing of multiple sensor updates with high throughput rates, prediction of moving objects, and on-line quality assessments, and demonstrate the feasibility of this approach by implementing the generic system architecture with laser scanners for object detection.

Parameter Identification of Ship Maneuvering Models Using Recursive Least Square Method Based on Support Vector Machines

Determination of ship maneuvering models is a tough task of ship maneuverability prediction. Among several prime approaches of estimating ship maneuvering models, system identification combined with the full‐scale or free‐ running model test is preferred. In this contribution, real‐time system identification programs using recursive identification method, such as the recursive least square method (RLS), are exerted for on‐line identification of ship maneuvering models. However, this method seriously depends on the objects of study and initial values of identified parameters. To overcome this, an intelligent technology, i.e., support vector machines (SVM), is firstly used to estimate initial values of the identified parameters with finite samples. As real measured motion data of the Mariner class ship always involve noise from sensors and external disturbances, the zigzag simulation test data include a substantial quantity of Gaussian white noise. Wavelet method and empirical mode decomposition (EMD) are used to filter the data corrupted by noise, respectively. The choice of the sample number for SVM to decide initial values of identified parameters is extensively discussed and analyzed. With de‐noised motion data as input‐output training samples, parameters of ship maneuvering models are estimated using RLS and SVM‐RLS, respectively. The comparison between identification results and true values of parameters demonstrates that both the identified ship maneuvering models from RLS and SVM‐RLS have reasonable agreements with simulated motions of the ship, and the increment of the sample for SVM positively affects the identification results. Furthermore, SVM‐RLS using data de‐noised by EMD shows the highest accuracy and best convergence. http://www.transnav.eu the International Journal on Marine Navigation and Safety of Sea Transportation Volume 11

A flexible framework for multisensor data fusion using data stream management technologies

Many applications use sensors to capture an image of the real world, which is needed for automatic processes. E. g. future driver assistance systems will be based on dynamic information about the cars environment, the cars state and the drivers state. Since there exists no single sensor that can sense the required information, different sensors like radar, video and eye-tracker are used. Typically some provide redundant information about the same real world entity, while others measure different things. Thus, the fusion of information from different sensors is necessary to get a consistent image of the real world. In most sensor fusion systems the sensor configuration is known a priori and the fusion algorithms are adapted for these sensor configurations. Thus, changing a sensor fusion system to enable it to process sensor readings from another sensor configuration is hardly possible or completely impossible. Since in development processes of automotive applications different sensor equipment and environmental requirements exist and change frequently a new approach for adapting sensor fusion systems is necessary. Hence, in this work a framework for sensor fusion systems will be developed that allows a flexible adaptation of fusion mechanisms. Due to real-time requirements of automotive applications and the flexibility of query processing technologies, data stream management technology will be used to develop a flexible framework for multisensor data fusion.

Prediction functions in bi-temporal datastreams

Modern datastream management system (DSMS) assume sensor measurements to be constant valued until an update is measured. They do not consider continuously changing measurement values, although a lot of real world scenarios exist that need this essential property. For instance, modern cars use sensors, like radar, to periodically detect dynamic objects like other vehicles. The state of these objects (position and bearing) changes continuously, so that it must be predicted between two measurements. Therefore, in our work we develop a new bitemporal stream algebra for processing continuously changing stream data. One temporal dimension covers correct order of stream elements and the other covers continuously changing measurements. Our approach guarantees deterministic query results and correct optimizability. Our implementation shows that prediction functions can be processed very efficiently.

Comparison and optimization of the parameter identification technique for estimating ship response models

Parameter identification techniques assorted from the system identification technology is a sufficient and commonly used approach for estimating the parameters of ship dynamic models. It is not tough to find an identification method to identify the parameters of linear or nonlinear ship dynamic models, but how to select a suitable parameter identification approach with high accuracy and low complexity for special cases is necessary to be studied. This contribution aims at determining a relatively suitable parameter identification method for estimation ship response models via selecting and comparing one intelligent method with the classic least squares method (LS) from a methodological point of view. Support vector machines (SVM) as an intelligent method is chosen because it is a kind of batch identification technique requiring no initial estimation of identified parameters. For well-confirming parameters in SVM, the artificial bee colony (ABC) algorithm instead of the empirical method is used to optimize the parameters in SVM. With the measurement zigzag test data from a scaled-model ship as training and verification samples, the maneuvering indices of ship response models are respectively identified using LS and SVM, and the verification of the identified models are sufficiently proceeded through comparing the prediction and measurement results. It is shown that the two different categories of ship response models are analytically and numerically consistent with each other. Comparison between the measured and predicted maneuvers demonstrates that SVM optimized by ABC algorithm is also an effective parameter identification technique.

Save maritime systems testbed

Abstract ‘Safe voyage from berth to berth’ — this is the goal of all e-navigation strains, driven by new technologies, new infrastructures and new organizational structures on bridge, on shore as well as in the cloud. To facilitate these efforts suitable engineering and safety/risk assessment methods have to be applied. Understanding maritime transportation as a sociotechnical system allows system engineering methods to be applied. Formal and simulation based verification and validation of e-navigation technologies are important methods to obtain system safety and reliability. The modelling and simulation toolset HAGGIS provides methods for system specification and formal risk analysis. It provides a modelling framework for processes, fault trees and generic hazard specification and a physical world and maritime traffic simulation system. HAGGIS is accompanied by the physical test bed LABSKAUS which implements a reference port and waterway. Additionally, it contains an experimental Vessel Traffic Services (VTS) implementation and a mobile integrated bridge enabling in situ experiments for technology evaluation, testing, ground research and demonstration. This paper describes an integrated seamless approach for developing new e-navigation technologies starting with virtual simulation based assessment and ending in physical real world demonstrations.