Search

Being difficult to scale poses great problems in multi-agent coordination. Multi-agent Reinforcement Learning (MARL) algorithms applied in small-scale multi-agent systems are hard to extend to large-scale ones because the latter is far more dynamic and the number of interactions increases exponentially with the growing number of agents. Some swarm intelligence algorithms simulate the release and utilization mechanism of pheromones to control large-scale agent coordination. Inspired by such algorithms, \textbf{PooL}, an \textbf{p}her\textbf{o}m\textbf{o}ne-based indirect communication framework applied to large scale multi-agent reinforcement \textbf{l}earning is proposed in order to solve the large-scale multi-agent coordination problem. Pheromones released by agents of PooL are defined as outputs of most reinforcement learning algorithms, which reflect agents' views of the current environment. The pheromone update mechanism can efficiently organize the information of all agents and simplify the complex interactions among agents into low-dimensional representations. Pheromones perceived by agents can be regarded as a summary of the views of nearby agents which can better reflect the real situation of the environment. Q-Learning is taken as our base model to implement PooL and PooL is evaluated in various large-scale cooperative environments. Experiments show agents can capture effective information through PooL and achieve higher rewards than other state-of-arts methods with lower communication costs., Comment: preprint

The effects of the Ni addition on the microstructures of Mg-2Gd-xNi (at.%, x = 0, 0.25, 0.5 and 1) alloys were systematically investigated. A series of two-phase Mg-Gd-Ni alloys consisting of long period stacking ordered (LPSO) phase and α-Mg were obtained. The LPSO phase presented block-like shape and was distributed between α-Mg dendrite arms. Its volume fraction was found to be proportional to the Ni content, ranging from 0 to ∼40% with the Ni content increasing from 0 to 1 at.%. After a heat treatment at 500 °C for 100 h, the LPSO phase experienced a structure transformation from 18R to 14H but was hardly dissolved into the α-Mg matrix, which evidenced a high thermal stability. The mechanism for the effects of the Ni addition on the formation of the LPSO phase was discussed.

Geospatial Web Services are data-oriented services, which include a variety of complex data models and metadata. Discovering the appreciate services with related geospatial datasets among a large number of available ones is a key task in the Geospatial Web Services domain. This paper proposes a peer-to-peer (P2P) based approach for discovering geospatial Web Services. We characterize the geospatial Web Services profile as a set of keywords including the metadata attributes, the minimum bounding rectangle (MBR) and the QoS parameters. Differing from the keywords based P2P Web Services discovery approaches, we use the MBR information to cluster and index the services into a kind of Peer R+ tree. With the tree, a P2P system can support complex queries containing partial keywords and spatial querying. The approach has been used in the Beijing Spatial Data Infrastructure project. We implement an peer-to-peer geospatial Web Services discovery system prototype that shows our approach facilitates complex service queries that contains spatial keywords.

GIS is a good spatial analysis tool, nevertheless, as the accumulation of spatial data, the functions that GIS offers are not enough. Spatial data mining, which can automatically discover implicit knowledge from spatial data, has recently received wide attention. However, the spatial data mining systems are not competent for preprocessing and presenting spatial data. Hence, the idea to integrate spatial data mining and GIS is straightforward. In this paper, we propose a novel method to integrate spatial data mining and GIS. This method is implemented with SPMML, an XML-based language that is extended from PMML. We build a prototype system with this method and it works well.

Outlier detecting is one of the most important data analysis technologies in data mining, which can be used to discover anomalous phenomena in huge dataset. Many literatures on spatial outlier detecting and time series outlier detecting have appeared, while the area of spatio-temporal outliers considering both spatial and temporal dimensions has still rarely been touched. Defining outliers in traditional dataset is more explicit because the data structure we need to focus on is very straightforward (e.g., a spatial point or a transaction record). However, it is much more difficult to give outlier a definite characterization in spatio-temporal lattice data, since there are so many data structures we can pay attention to. With the aim of detecting useful and meaningful outliers in climate dataset, we introduce a formalized way to define outliers in spatio-temporal lattice data, in which the importance of clarifying basic data structure (we call it basic element in our paper) is stressed. As a case study, we define two kinds of spatio-temporal outliers based on a global climate dataset, according to the three aspects we propose in defining an outlier. The introduction of basic element and the formulation of outlier definition process make it easier and clearer to define meaningful outliers. Thus outlier detecting in spatio-temporal lattice data will provide us with really interesting and useful knowledge.

Spatial data mining is a promising technique that deals with extraction of implicit knowledge or other interesting patterns from large amount of spatial data. Though most data mining systems work with data stored in flat files or operational database, it has been recognized that mining in a data warehouse usually result in better information. Because data are usually cleansed before they are stored into data warehouse. Furthermore, data warehouse provides data with different levels of summarization for the clients, which will lead to fruitful data mining. However, current techniques of data warehouse can not handle spatial data well. Both dimensions and measures in the data model of data warehouse are nonspatial data. In this paper, we propose a new data model called spatial data cube for data warehouse. Spatial data cube supports both spatial and nonspatial data. We also introduce how to construct a spatial data cube that can answer queries efficiently by selective materialization. We believe that the spatial data cube can provide better support for spatial data mining.

The location technologies, such as GPS and telegraphy, are producing more and more data of moving objects. Spatio-temporal database is needed to manage these data, so as to solve the problems in spatio-temporal applications. However, there are few prototypes of spatio-temporal database systems yet. This is because attention is always paid on some specified aspects not the whole system. In this paper, we present the design and implementation of a spatio-temporal database prototype for managing moving objects. A Trajectory Model is proposed in this paper, where the moving objects' movements in 2D plane over time are treated as trajectories in the 3D spatio-temporal space. Referring to R tree and the trajectory oriented indexing method TB tree, a new indexing method: S-TB tree is proposed in the prototype. Distance based queries and trajectory topology based queries are all supported by the prototype. The distance based queries includes point queries, range queries. This paper also presents different strategies to process these queries. Finally, two application cases are presented: traffic control and historical queries of the storms on Earth.

GIS Web service turns out to be one of the new directions of the paradigm of GIS, due to the popular use of the Internet and the dramatic progress of telecommunications technology. However, focused on the description of the interface of functions by Web service description language (WSDL), the general Web service technologies are not sufficient to model GIS Web service with data-oriented characteristic, especially those data provider services and data transform services. In this paper, we propose geographic-Web service description language (abbr. G-WSDL) which enhances the ability of describing such data-driven services by extending the current WSDL with service metadata. Two application scenarios will be presented at last to show its strength in automatic GIS Web service discovery and dynamic services combination.

Spatial sampling is a significant part of query processing in spatial database. In this paper, an adaptive data-driven sampling for selectivity estimation is proposed in spatial database. This technique presents an efficient sampling for spatial data, especially for two-dimensional line and polygon data, and it make the sample size fit the limit of time and memory, or a user-defined parameter. The data-driven sampling technique is compared with various techniques on different type of datasets in our experimental study, and it out outperforms the other techniques over a broad range of query workloads and datasets.