SDMIS 2021 Abstracts

Area 1 - SDMIS

Full Papers
Paper Nr: 1

Efficient Secure Communication for Distributed Multi-Agent Systems


Davide Costa, Daniel Garrido and Daniel C. Silva

Abstract: The use of multi-agent systems has been increasing and with it the need to improve communication performance and to make it secure. In a system with hundreds of agents, in which their reaction must be fast, it is essential to ensure low latency and high message throughput. These agents can work in a cooperative or competitive environment and, especially in the last, the absence of secure communications opens the possibility for malicious agents to intercept and/or change the content of the messages. This paper explores alternatives to the Message Transport Service as described in the Foundation for Intelligent Physical Agents architecture for multi-agent systems, namely using message-oriented middleware. It also introduces a new component to the typical multi-agent system architecture, the certification authority service. This component is responsible for creating certificates that platform agents can use to ensure their identity and communicate safely. This architecture also manages external agents distribution across several machines, similar to a federated environment, making the system more suitable for computationally demanding scenarios. The architecture was tested on an existing simulation platform, showing very good results.

Paper Nr: 4

Vector Quantization to Visualize the Detection Process


Kaiyu Suzuki, Tomofumi Matsuzawa, Munehiro Takimoto and Yasushi Kambayashi

Abstract: One of the most important tasks for drones, which are in the spotlight for assisting evacuees of natural disasters, is to automatically make decisions based on images captured by on-board cameras and provide evacuees with useful information, such as evacuation guidance. In order to make decision automatically from the aforementioned images, deep learning is the most suitable and powerful method. Although deep learning exhibits high performance, presenting the rationale for decisions is a challenge. Even though several existing decision making methods visualize and point out which part of the image they have considered intensively, they are insufficient for situations that require urgent and accurate judgments. When we look for basis for the decisions, we need to know not only WHERE to detect but also HOW to detect. This study aims to insert vector quantization (VQ) into the intermediate layer as a first step in order to show HOW to detect for deep learning in image-based tasks. We propose a method that suppresses accuracy loss while holding interpretability by applying VQ to the classification problem. The applications of the Sinkhorn–Knopp algorithm, constant embedding space and gradient penalty in this study allow us to introduce VQ with high interpretability. These techniques should help us apply the proposed method to real-world tasks where the properties of datasets are unknown.

Paper Nr: 5

Advances in Hybrid Evolutionary Algorithms for Fuzzy Flexible Job-shop Scheduling: State-of-the-Art Survey


Mitsuo Gen, Lin Lin and Hayato Ohwada

Abstract: Flexible job shop scheduling problem (FJSP) is one of important issues in the integration of research area and real-world applications. The traditional FJSP always assumes that the processing time of each operation is fixed value and given in advance. However, the stochastic factors in the real-world applications cannot be ignored, especially for the processing times. In this paper, we consider FJSP model with uncertain processing time represented by fuzzy numbers, which is named fuzzy flexible job shop scheduling problem (F-FJSP). We firstly review variant FJSP models such as multi-objective FJSP (MoFJSP), FJSP with a sequence dependent & set time (FJSP-SDST), distributed FJSP (D-FJSP) and a fuzzy FJSP (F-FJSP) models. We secondly survey a recent advance in hybrid genetic algorithm with particle swarm optimization and Cauchy distribution (HGA+PSO) for F-FJSP and hybrid cooperative co-evolution algorithm with PSO & Cauchy distribution (hCEA) for large-scale F-FJSP. We lastly demonstrate the HGA+PSO and hCEA show that the performances better than the existing methods from the literature, respectively.

Paper Nr: 6

Trading Agent Competition with Autonomous Economic Agents


David Minarsch, Seyed Ali Hosseini, Marco Favorito and Jonathan Ward

Abstract: We provide a case study for the Autonomous Economic Agent (AEA) framework; a toolkit for the development and deployment of autonomous agents with a focus on economic activities. The use case is the trading agent competition (TAC). It is a competition between autonomous agents with customisable strategies and market parameters. The competition is facilitated by the AEA framework’s native support for decentralised ledger technologies, i.e. permissionless blockchains and smart contract functionality, for immutable transaction recording and trade settlement. We provide an open-source implementation, study the result of the competitions we ran, and compare it to theoretical results in the economics literature. We conclude by discussing its real-world applications in crypto-currency, digital assets and token trading.

Short Papers
Paper Nr: 2

Comprehensive Transcriptional Analysis Reveals Gene-specific Transcriptional Variations in a Seed Plant, Arabidopsis thaliana


Kohei Negishi and Kengo Morohashi

Abstract: The multicellular biological organism comprises a number of cells connected, while each cell independently works. It seems to have a system to orchestrate a number of cells, like a parallel multi-agent intelligent system. In such a biological system, gene expression of even identical genes within homogeneous cell populations is varied due to a stochastic fluctuation of the transcriptional process. This gene expression variation (GEV) is observed in development, cell differentiation, and environmental responses. Although the GEV has been generally reported, a gene-specific GEV remains unclear. Using publicly available genome-wide gene expression data from a model plant, Arabidopsis thaliana, we successfully identified two groups of genes whose GEVs demonstrated consistently high and low. Analysis of 632 experimental conditions derived from more than 1,300 microarrays revealed that 65 and 296 genes had high and low GEVs, respectively. We named genes with the high GEV DOTABATA (DTBT), which means “romping about” in Japanese, and genes with the low GEV PISHIPASHI (PSPS), which means “over-discipline” in Japanese. Gene function enrichment analysis resulted that DTBT genes significantly enriched stress response genes. Our results suggest a gene-specific GEV, and the regulation of GEV would be involved in biological processes.

Paper Nr: 3

Multi-class Motor Imagery EEG Classification using Convolution Neural Network


Amira Echtioui, Wassim Zouch, Mohamed Ghorbel, Chokri Mhiri and Habib Hamam

Abstract: Electroencephalogram (EEG) signals based on Motor Imagery (MI) are a widely used form of input in Brain Computer Interface (BCI). Although there are several ways to classify data, a question remains as to which method to use in EEG signals based on motor imagery. This article presents an attempt to reach the best classification method based on deep learning methods by comparing two models: Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM), on the same basic data set. The BCI Competition IV dataset 2a was used as the base dataset to test the two classification methods. Experimental results show that the proposed CNN model outperforms the LSTM model, with an accuracy value of 74%, and other state-of-the-art methods.