III Jornada MOSI-AGIL

 III Jornada sobre Modelado Social de Inteligencia Ambiental Aplicado a Grandes Instalaciones

 

Director: Alberto Fernández Gil

Fecha: 24 de octubre de 2017

Lugar: Universidad Rey Juan Carlos. Salón de Grados del Departamental II. Campus Móstoles.

Cómo llegar: http://www.urjc.es/universidad/campus/campus-de-mostoles/564-situacion-planos-campus-mostoles 

Organizadores

  • Holger Billhardt. Universidad Rey Juan Carlos
  • Alberto Fernández Gil. Universidad Rey Juan Carlos
  • Sascha Ossowski. Universidad Rey Juan Carlos
  • Carlos Ceacero. Universidad Rey Juan Carlos
  • Álvaro Carrera. Universidad Politécnica de Madrid
  • Ramón Alcarria. Universidad Politécnica de Madrid
  • Jorge J. Gómez. Universidad Complutense de Madrid

Contacto: Contactar al organizador

Alumnos:

  • Reconocimiento Académico de Créditos (para alumnos URJC): Se reconocerán 0.2 (por confirmar) créditos por la asistencia a la jornada a los alumnos de Grado de la URJC
  • Parte de las Actividades Formativas del Programa de Doctorado en Tecnologías de la Información y las Comunicaciones de la URJC
  • Indistintamente de la procedencia, se hará un certificado de asistencia a las jornadas

Temática:

Dada la tendencia de consolidación de grandes grupos de población, se hace necesario disponer de medios fiables para gestionar las instalaciones que los alojarán en algún momento (grandes superficies comerciales, aeropuertos, estaciones de tren, polideportivos y similares). Existen ya expertos que las diseñan y las ejecutan, pero faltan las herramientas y conocimientos para justificar de forma más científica los límites de las instalaciones, para evaluar las situaciones (siempre se identifica una crisis cuando ya es demasiado tarde) y para tomar decisiones.

El objetivo de esta tercera jornada sobre Modelado Social de Inteligencia Ambiental Aplicado a Grandes Instalaciones es crear  nuevamente un foro común para el intercambio de ideas y presentar avances en el tema de gestionar el comportamiento de personas en grandes instalaciones. En esta ocasión se presentan varios avances en tema de simulación como en sistemas de guiado de personas.

Al igual que las anteriores ediciones, la jornada se enmarca dentro del conjunto de actividades del Programa de I+D “Modelado Social de Inteligencia Ambiental Aplicado a Grandes Instalaciones (MOSI-AGIL-CM, S2013/ICE-3019), financiado por la Comunidad de Madrid y en el que el Grupo de Inteligencia Artificial de la URJC actúa como coordinador. Además de la URJC, en el programa participan dos grupos de investigación de la UPM (GSI y GISAI) y UCM (GRASIA).

Ediciones anteriores: 2015, 2016

Programa: 

10:00-10:10 Apertura de la Jornada

10:10-11:30 Presentaciones

  • Marin Lujak (IMT Lille Douai, Francia): On Shoppers' Route Guidance in Congested Smart Hypermarkets (presentación)
  • Marlon Cárdenas (Universidad Complutense de Madrid): Virtual Development of a Presence Sensor Network Using 3D Simulations (presentación)
  • Alberto Fernández (Universidad Rey Juan Carlos): Agile Smart Building Evacuation using Event Processing and Semantic Technologies (presentación)

11:30-12:00 Descanso – café

12:00-13:30 Presentaciones

  • Borja Bordel (Universidad Politécnica de Madrid): Trust, reputation and ambient intelligence: detecting malicious components in pervasive sensing systems and smart environments (presentación)
  • Álvaro Carrera Barroso (Universidad Politécnica de Madrid): Extending a social simulation framework to achieve simulation as a service (presentación)
  • Felix Manuel Mellado Romero (Universidad Rey Juan Carlos): Componente software para la generación de rutas de evacuación de edificios (presentación)

13:30-14:00 Discusión/Debate/Colaboraciones/Propuestas de TFGs/TFMs para alumnos

14:00 Clausura

 

Resúmenes Presentaciones:

  • Marin Lujak (IMT Lille Douai, Francia): On Shoppers' Route Guidance in Congested Smart Hypermarkets. In this talk, we consider the problem of route guidance for shoppers in congested hypermarkets equipped with smart space technologies. This is both an actual problem due to the time lost in waiting queues in peak hours and a highly computationally complex problem due to dynamically changing congestion conditions, the size and complexity of hypermarkets, and the presence of a multitude of shoppers with different shopping constraints. We model the hypermarket by a network of communicating infrastructure agents, each one monitoring the conditions on a separate physical area of a hypermarket. Each shopper is modeled as an agent installed on a shopper’s app that, by interacting with other shoppers and hypermarket infrastructure agents, dynamically updates its efficient and effective shopping route. Individual routes are computed through a proposed distributed shopping route optimization algorithm that distributes information and computation in such a multi-agent system. Each shopper agent resolves the pick sequencing problem, i.e., given a shopper’s list, the items’ locations are sequenced in the route proposed to a shopper so that the overall traveling time is minimized considering congestion in real-time. The complexity of this approach depends on its mathematical modelling. We discuss several models and propose the most scalable one. We experiment the proposed algorithm in simulations comparing it with a centralized one and a greedy tour algorithm. Our preliminary results show that the proposed approach scales well and provides efficient shoppers’ routes
  • Alberto Fernández (Universidad Rey Juan Carlos): Agile Smart Building Evacuation using Event Processing and Semantic Technologies. In this work we consider the problem of evacuation of smart buildings in case of emergencies. In particular, we present a proposal for an evacuation guidance system that provides individualised evacuation support to people in case of emergencies. The system uses sensor technologies and Complex Event Processing to obtain information about the current situation of a building in each moment. Using semantic Web technologies, this information is merged with static knowledge (special user characteristics, building topology, evacuation knowledge) in order to determine feasible evacuation routes for each user. In particular, we look for routes with high probability of easy re-routing in case of contingencies. For this, we use the notion of agility for selecting the most appropriate evacuation route for each person out of a possible set of candidates.
  • Álvaro Carrera Barroso (Universidad Politécnica de Madrid): Extending a social simulation framework to achieve simulation as a service. In this talk, we explore the extension of an existing social simulation framework focused on building evacuation, such as UbikSim, to enable the development of a simulation service implemented as a web service. For emergency planning, it is important to evaluate different potential situations with different kinds and size of emergency. That evaluation can be simplified applying simulation techniques for generating those situations in a virtual environment and for varying many parameters as considered to create realistic simulations in a specific building. This talk shows how UbikSim desktop tool is extended to define and configure the environment simulation through an API or a configuration file. This extension enables a wide variety of applications to collect simulation results as a service and allows third-party autonomous decision support systems to consume those resulting data, such as the number of injured people during the evacuation, progression of fire in the building, crowded locations, etc, to make decisions about the emergency plan of a building.
  • Felix Manuel Mellado Romero (Universidad Rey Juan Carlos): Componente software para la generación de rutas de evacuación de edificios. En esta charla se presentará el desarrollo de un componente software adaptable que permite generar rutas de evacuación individualizadas para las personas que se encuentran en un edificio. El componente recibe la estructura del edificio y su estado en un momento determinado (incluyendo localización de las personas) y genera las rutas recomendadas hacia una salida para cada persona que se encuentre en su interior. El componente es fácilmente modificable para incluir diferentes algoritmos de cálculo de rutas.
  • Marlon Cárdenas (Universidad Complutense de Madrid): Virtual Development of a Presence Sensor Network Using 3D Simulations. Testing the control and deployment of large networks of sensors and actuators is a complex and expensive task. This paper presents a 3D simulation tool that facilitates testing and measuring this kind of systems in a virtual environment, which alleviates the costs of doing these tasks in a physical setting. This is illustrated with an example of how a presence detection system can be designed to monitor the behavior of a crowd under different stimulus.
  • Borja Bordel (Universidad Politécnica de Madrid): Trust, reputation and ambient intelligence: detecting malicious components in pervasive sensing systems and smart environments. Nowadays, ambient intelligence (AmI) solutions include very complex architectures which involve a great amount of components (such as devices, services, execution engines, etc.). This complexity facilitates the appearance of malicious components; those which provide uncertain data, services or information. In general, AmI systems try to merge very different devices and other components. Thus, low‐level information must be collected, transformed, aggregated and translated various times before being sent to the high‐level final applications. However, none meta‐information about the underlying hardware platform is provided to the high‐level layers. Therefore, final applications have a very limited knowledge about the system which provides them with the operation data. In this context, new solutions to calculate important concepts such as the trust level or the component reputation based only on the available information at high‐level are necessary. In this talk a statistical framework for knowledge discovery in order to estimate the uncertainty level associated with the received operation data by a certain application is presented. Additionally, these results are used as input in a reputation model focused on locating the malicious components. Thus, if possible, final applications may discard information from these components.

 

 Financiación:

 

Programa MOSI-AGIL-CM (S2013/ICE-3019), financiado por: