Programma

Program ICIS Final Event January 26 2010

 10.00Opening, Prof. Dr. H.J. van den Herik
   
 Track 1Track 2Track 3 
 10.30Highlights & VisionResearchValorization projects
 Common Situation Awareness                    Collaborative Decision MakingTraffic
 Chair & intro: Frans GoenCo-chair & intro: Martijn NeefChair: Dick Fikkert
 Method for Designing Networking Adaptive Interactive Hybrid Systems
by Leon Kester



Online, data-efficient reinforcement learning
by Lucian Busoniu

 

PAGE
by Gijs Withagen (Technolution)
 Enhanced Situation
Awareness in Crisis Management: a probabilistic approach
by Patrick den Oude

Give me autonomy!
by Matthijs Amelink

i-Catcher
by Hans Lammers(Logica)
    
 11.30Integrative ArchitecturesEnhanced Situational AwarenessDecision Support
 Chair & intro: Rob MeijerCo-chair & intro: Leon KesterChair: Hans Keus
 GRAMM
by Sander van Splunter & Bernard van Veelen


Cooperative Learning
Algorithm for Control Problems in the Continuous Domain
by Jelmer van Ast

Promedas
by Wim Wiegerinck (SNN)
 SADDOG
by Sorin Iacob
Autonomy and Coordination
by Bob van der Vecht
Bonaparte
by Wim Wiegerinck
(SNN)
  Multi-robot Exploration with Limited Communication in the RoboCup Rescue
by Arnoud Visser
 
    
 12.30Lunch & demo'sLunch & demo'sLunch & demo's
    
 14.00Multi modal interactionICIS ArchitectureActor Agent Communities
 Chair & intro: Lou BovesCo-chair & intro: Bernard van VeelenChair: Kees Nieuwenhuis
 CHIM demostrator
by Job Zwiers
Dynamic scheduling problems call for adaptive heuristics
by Jeroen de Jong

DEIN
by Gregor Pavlin (Thales)
  A Semantic Model for Complex Computer Networks – The Network Description Language
by Jeroen van der Ham
MAS@NS: Train Driver Rescheduling
by David Mobach
    
 15.00Collaborative decision makingComputer Human Interaction ModelingHealthcare & Distributed Sensor Networks
 Chair & intro: Robert BabuskaCo-chair & intro: Lynn PackwoodChair: Patrick Storms
 Actor-Agent Teaming: The Human (F)Actor
by Niek Wijngaards
Local Danger Warnings for Drivers: The effect of presentation on Driver Reaction
by Yujia Cao


Fundamentals of CHAP applied in SenZorg
by Duco Ferro (Almende)

 The Role of Sensemaking and Information Management in Crisis Response
by Willem Muhren
Multimodal recognition of emotions
by Dragos Datcu
SAIL
by Jan Katgerman
(Rups)
    
 16.00Final speech Dr. C.H.M. Nieuwenhuis & presentation ICIS Scientific Book
 16.30Drinks

 

Presentation abstracts

Track 1: Highlights & Vision

Method for Designing Networking Adaptive Interactive Hybrid Systems
(Leon Kester)
How to integrate Enhanced Situational Awareness in Actor Agent Communities for various application domains, including a design method and a vision to the future.

Enhanced Situation Awareness in Crisis Management: a probabilistic approach
(Patrick den Oude)
A probabilistic approach for critical state estimation in the context of the crisis management domain focusing on the detection of toxic gases and gas leak localization using probabilistic networks.

GRAMM
(Sander van Splunter & Bernard van Veelen)
Management in crisis response requires continuous adaptation, for crisis situations are highly  dynamic. Crisis response in general involves multiple parties, each with their own autonomy and capabilities, leading to differentiations in structure, goals andstrategies, and constraints for cooperation. A crisis management system needs to support distributed and continuous adaptation on different levels of organisation, in a reliable fashion, ensuring at least some minimal level of service for every defined task. A generic architecture is developed that allows for multi-level reasoning on the global impact of local adaptations, and is applied to coordination of fire-fighters in multiple safety regions.

SADDOG
(Sorin Iacob)
An approach to define and implement an end-to-end security solution for data distributed systems aiming at protecting the confidentiality and integrity of data objects for their entire lifetime.

CHIM Demonstrator
(Job Zwiers)
Human-system interaction is one of the challenges in scenarios where there is a lot of information exchange, within a short period of time, with high workloads and under stressful conditions. The Chim demonstrator uses modalities like pen input, speech,  gestures, and facial expressions, as inputs for a robust dialogue system, and multimodal output generation. The goal of the system is that is able to cope with ambiguous and unreliable information coming from humans like firemes, police or medics,  that are involved in  fighting a disaster situation in a tunnel.

Actor-Agent Teaming: The Human (F)Actor
(Niek Wijngaards)
Humans and agents working together is challenging: their joint results do not constitute a-priori performance improvement. Not only do agents need team-oriented behaviour, they also need understanding of humans. In particular, how humans partake in decision making, alone and together.

The Role of Sensemaking and Information Management in Crisis Response
(Willem Muhren)
Crisis environments are characterized by various types of information and communication challenges which complicate the work that responding organizations have to do. We have conducted case studies in different types of crisis situations, in which we investigated in what way actors manage and process information, how it helps them to make sense of what is going on, and how information systems can support them in doing this.

Track 2: Research

Cooperative Learning Algorithm for Control Problems in the Continuous Domain
(Jelmer van Ast)
Algorithms, theoretical analysis and experimental results of multiple agents cooperating to learn optimal control policies, based on Ant Colony Optimization

Autonomy and Coordination
(Bob van der Vecht)
When designing interactive collaborative systems, one needs to balance between autonomy of individuals and coordination of group behavior. In this presentation we discuss the relation between autonomy and coordination and we propose a system description that includes both perspectives.

Online, data-efficient reinforcement learning
(Lucian Busoniu)
Reinforcement learning is an algorithmic paradigm inspired from human learning processes. In this presentation, we describe several data-efficient methods for online reinforcement learning, together with their application to the control of physical systems, including a robotic goalkeeper arm.

Give me autonomy!
(Matthijs Amelink)
Illustrative example of a step by step analysis of a complex human-machine system with the intent to fully automate it. Question are raised what autonomy means and if how it would be achievable. By looking into the system through systematic work domain analysis, it is broken down into control problems with representations for control and decision making. Which of these control problems can be automated and which cannot is discussed using a summary of the complete Work Domain Analysis.

Multi-robot Exploration with Limited Communication in the RoboCup Rescue
(Arnoud Visser)
An overview of the progress made in the RoboCup Rescue League, concerning cooperation inside teams of robots which has to perform a search and rescue mission, with a focus on the work of the Amsterdam Oxford Joint Rescue Forces.

Dynamic scheduling problems call for adaptive heuristics
(Jeroen de Jong)
How to deal pragmatically with scheduling problems where new jobs arrive constantly making current plans obsolete? This talk dives into scheduler adaptiveness on short and longer term."

A Semantic Model for Complex Computer Networks – The Network Description Language
(Jeroen van der Ham)
About a formal semantic model, the Network Description Language (NDL), which provides a way to describe complex computer networks in a meaningful, and interoperable way. These descriptions are used to create a distributed information model for describing multi-layer, multi-domain computer networks. This model can be used to do pathfinding, support engineers in managing the network, and provide applications insight into the possibilities of the network.

Local Danger Warnings for Drivers: The effect of presentation on Driver Reaction
(Yujia Cao)
Local danger warning is an important function of Advanced Driver Assistance Systems (ADAS) to improve the safety of driving. Besides directly sensing the environment to detect danger, recent advances in inter-vehicle communication technology (e.g. wireless ad-hoc networks car-2-car communication) further allow the exchange of information between cars. This enables a much wider application of local danger warnings, as drivers can be alerted to approaching danger that is not yet visible. A crucial part of successful danger avoidance is the user interface – the presentation of the warning to the driver. In this talk, I present a user study investigating various warning presentations using a scenario of emergent road obstacles. Two presentation factors were selected: modality and level of assistance (AS). In accordance with the ISO usability model, a total of 6 measurements were derived to assess the effectiveness and efficiency of the warnings and the drivers’ satisfaction. Results of this study have implications to the design of local danger warnings, and the design of in-vehicle warnings in general.

Multimodal recognition of emotions
(Dragos Datcu)
Multimodal recognition of emotions has been lately studied by several research groups worldwide. In our research we focussed on different approaches for automatic recognition of emotions by considering vision and speech data. The modalities have been investigated separately and together in order to identify which models lead to better recognition results. The extraction of specific indicators of emotions from speech audio data has been realized using prosodic features and classification models like Gentle Boost and hidden Markov models. The facial expressions are identified using static and dynamic models. In either case we have used parameter facial representations from adapted local binary patterns, Viola&Jones features, geometric and optical flow features. Our facial model is based on Active Appearance model. We were able to develop a running prototype which will be demonstrated.

Track 3: Valorization

PAGE
(Gijs Withagen)
Technolution developed a new type of system to guide cars to available parking places throughout the city. The new agent-based system will combine information from several information systems to be able to guide traffic more intelligently. In the project is cooperation between Parkeergeleiding Rotterdam, Technolution and the Municipality of Rotterdam.

SAIL
(Jan Katgerman)
The SAIL Depth sounding project developed a new method to create up to date depth maps of inland waterways. For testing in the area of Rotterdam inland ships will intelligently gather and share measurements using innovative techniques from the ICIS project. Shippers and waterway authorities will receive better and cheaper information tailored to their specific needs and situation.

PROMEDAS
(Wim Wiegerinck)
Promedas is medical diagnostic decision support system that uses probabilistic inference. At the moment, Promedas covers thousands of diagnoses, symptoms and their relationships  Its practical value has been demonstrated and evaluated in the Universitair Medisch Centrum Utrecht (UMCU).

BONAPARTE
(Wim Wiegerinck)
Bonaparte is a decision support system to improve forensic institute’s missing person screening and matching for victim identification based on DNA profiles, e.g. in crisis situations. The system uses Bayesian network modeling and inference methods.

Fundamentals of CHAP applied in SenZorg
(Duco Ferro)
CHAP : an architecture for developing actor-agent communities.  We discuss CHAPs basic principles (e.g., improving the self-organization amongst people) and its application to the health care domain (SenZorg).

i-Catcher: Information Led Policing Crisis Control in Public Spaces
(Hans Lammers)
The aim of i-Catcher is to setup a collaborative, event driven, environment that supports the (automated) decision making process in a surveillance environment or emergency situations. The environment is based on doctrines from the defence industry. Key concepts are "common situational picture" and "mission critical C2". 
The environment uses different sensor technologies, such as intelligent video processing, sniffers, biometrics or agression detection. 
The environment will benefit any organization that has a sensor based surveillance assignment, while providing measures to guarantee the right of privacy of each individual in the public space. i-Catcher is a value proposition that is developed by Logica in close cooperation with different product suppliers.

DEIN
(Gregor Pavlin)
Dynamic Expertise Integration Networks (DEIN) is a novel tool for high-quality situation assessment validated in the crisis management. DEIN facilitates networking between experts and machines serving professionals in the field, such as fire fighter commanders, with the information on various crucial aspects. There is a strong involvement of Milieudienst Rijnmond (DCMR) in the project.

MAS@NS: Train Driver Rescheduling
(David Mobach)
The railway operations of Netherlands Railways (NS) are based on an extensive planning process. In particular, crew scheduling is a complex process. During operations plans have to be updated continuously in order to deal with delays of trains and larger disruptions of the railway system. Currently, these rescheduling operations are carried out by teams of experts, without the aid decision support systems. In recent years, the limits of this approach have become apparent. We present a multi-agent system aimed to support dispatchers in rescheduling of train drivers.

Demonstrators

SAIL
(Jan Katgerman)
The SAIL Depth sounding project developed a new method to create up to date depth maps of inland waterways. For testing in the area of Rotterdam inland ships will intelligently gather and share measurements using innovative techniques from the ICIS project. Shippers and waterway authorities will receive better and cheaper information tailored to their specific needs and situation.

PROMEDAS
(Wim Wiegerinck)
Promedas is medical diagnostic decision support system that uses probabilistic inference. At the moment, Promedas covers thousands of diagnoses, symptoms and their relationships  Its practical value has been demonstrated and evaluated in the Universitair Medisch Centrum Utrecht (UMCU).

BONAPARTE
(Wim Wiegerinck)
Bonaparte is a decision support system to improve forensic institute’s missing person screening and matching for victim identification based on DNA profiles, e.g. in crisis situations. The system uses Bayesian network modeling and inference methods.

DEIN
(Gregor Pavlin)
Dynamic Expertise Integration Networks (DEIN) is a novel tool for high-quality situation assessment validated in the crisis management. DEIN facilitates networking between experts and machines serving professionals in the field, such as fire fighter commanders, with the information on various crucial aspects. There is a strong involvement of Milieudienst Rijnmond (DCMR) in the project.

MAS@NS: Train Driver Rescheduling
(David Mobach)
The railway operations of Netherlands Railways (NS) are based on an extensive planning process. In particular, crew scheduling is a complex process. During operations plans have to be updated continuously in order to deal with delays of trains and larger disruptions of the railway system. Currently, these rescheduling operations are carried out by teams of experts, without the aid decision support systems. In recent years, the limits of this approach have become apparent. We present a multi-agent system aimed to support dispatchers in rescheduling of train drivers.

CHIM Demonstrator
(Job Zwiers)
Human-system interaction is one of the challenges in scenarios where there is a lot of information exchange, within a short period of time, with high workloads and under stressful conditions. The Chim demonstrator uses modalities like pen input, speech,  gestures, and facial expressions, as inputs for a robust dialogue system, and multimodal output generation. The goal of the system is that is able to cope with ambiguous and unreliable information coming from humans like firemes, police or medics,  that are involved in  fighting a disaster situation in a tunnel.

Research Simuation Kit (ReSK) (TRT)
(dr. Niek Wijngaards)
The RESK experimentation and simulation environment has been used in the SEAT subproject to conduct experiments to compare the performance of actor-only teams (consisting of six human team members) and actor-agent teams (consisting of three human and three agent team members) in an incident management situation. The simulation of a simplified environment in which both actors and agents are embodied (via similar ‘avatars’) and have to rescue victims and extinguish fires was deliberately engineered to focus on assessing teamwork and team performance. The demonstration at the ICIS Final Event lets the visitors participate in a simple actor-agent team.

Petrophysical Decision Support (SNN)
(Wim Wiegerinck)
This is a demonstrator for a petrophysical decision support system that helps to interpret measurement data acquired in boreholes and to find the most useful further measurement steps.
The system is based on petrophysical domain expertise, which is modeled with a  Bayesian network.
The demonstrator has been developed in collaboration with SHELL Exploration & Production.

Personal Intelligent Traveling Assistant
(Prof. drs. dr. L.J.M.Rothkrantz)
Nowadays, especially in the Netherlands, the road network is flooded by cars. In the rush hours a total of 200-300 kilometers of traffic jam is reported every day. To solve that problem dynamic routing can be used.
Most of the currently available route planners take no dynamic traffic information into account. They route car drivers from starting point to destination along the shortest route using some shortest path algorithm (Dijkstra). In rush hours the shortest path in distance is usually not equivalent to the shortest path in time. Dynamic routing is needed which takes into account current and tot be expected traffic congestion, incidents and other forms of delay. To assess the current/future traffic speed, special monitoring systems (MONICA) have been installed along the highways to compute current/future traffic speed. An alternative is to track car drivers using a mobile phone. At regular times they send an SMS with a timestamp and current positions. Based on that information dynamic routing algorithms can be used to route car drivers in the shortest time from A to B.

The Personal Intelligent Traveling Assistant (PITA) provides dynamic routing service to car drivers. They send via a mobile device a request to a server and a shortest route will be computed for very individual car driver. During his route the car driver send at regular times his position. This position is needed as input for the dynamic routing algorithm and to monitor the car driver. The success of the system depends of the amount of users and the quality of the used algorithms.
At this moment there is a running prototype available based on a dynamic version of the Dijkstra algorithms and the Ant Based Control algorithms.
A simulation of the PITA system will be demonstrated.

i-Catcher: Information Led Policing Crisis Control in Public Spaces
(Hans Lammers)
The aim of i-Catcher is to setup a collaborative, event driven, environment that supports the (automated) decision making process in a surveillance environment or emergency situations. The environment is based on doctrines from the defence industry. Key concepts are "common situational picture" and "mission critical C2". 
The environment uses different sensor technologies, such as intelligent video processing, sniffers, biometrics or agression detection. 
The environment will benefit any organization that has a sensor based surveillance assignment, while providing measures to guarantee the right of privacy of each individual in the public space. i-Catcher is a value proposition that is developed by Logica in close cooperation with different product suppliers.

Urban Search & Rescue
(UvA)
The Urban Search & Rescue demonstrator will be multirobot control software developed by the Amsterdam Oxford Joint Rescue Forces (www.jointrescueforces.eu).
Amsterdam and Oxford were interested in issues such as world modeling based on incomplete and inaccurate measurements and the coordination of the actions of multiple agents (such as robots) to achieve common goals.
It is important that these research issues are scaled up towards real-world applications with direct relevance to society. An application with significant potential is the use of mobile robots for search and rescue missions after a disaster (for instance in a contaminated area).

The Amsterdam Oxford Joint Rescue Forces participated in the RoboCup Rescue Simulation Virtual Robot competition. In this competition a devastated area has to be explored for victims with a team of robots controlled by a single operator. The robot team should accurately explore large areas with difficult mobility. The competition is based on a physics-based, realistic simulation on the sensory and actuation level of the robot. Standardized interfaces make a transparent migration of code between real robots and their simulated counterparts possible.

eCall
(LogicaCMG)
eCall is a system which will be installed in every new vehicle as off 2011. In case of a severe car crash, eCall sends an emergency message to 112 with the location of the accident and the nature. eCall will save lives , because the emergency services can respond quicker. They have the most relevant information at an instant.

eCall is a typical environment in which various organizations will have to collaborate technically. The in-vehicle system can be originated from any car manufacturer. The wireless communition (GSM) is often owned by a private operator. The 112 system is typically an high availability and fault tolerant system. Behind 112 there are the systems of police , fire brigade and ambulance.  So there is a chain of systems which exchange information with eachother. It is unlikely that all systems have a single a single architecture. Therefore it is a ideal environment for collaborative systems where each system can have  its own architecture, brand and ownership.

In the ICIS project Logica developed a demonstrator of eCall which shows the chain of collaborative systems. It consist of an on board unit which is based on a open services environment of OSGI/Java. Among all other in car functions, eCall is one of the Java bundles which triggers the emergency message. The emergency message is send over the GSM network to the simulated 112 BackOffice. This BackOffice is also used by Rijkswaterstaat for monitoring incident along the highway (IM4U).

Tunnel Operator Trainer
(TNO)
Tunnel Operator Training with a Conversational Agent-Assistant.
A tunnel operator monitors and regulates the flow of traffic inside a tunnel. At the ICIS final event, we present an automated training system for tunnel operator training. The system employs a conversational agent which supports the operator’s situation assessment tasks. The agent exhibits peer behavior which is unobtrusively directed by didactic strategies.

WILLEM: a Wireless InteLLigent Evacuation Method
(TNO)
A system for dynamic evacuation routing in buildings, using a wireless sensor network. Dynamic evacuation routing is the process of dynamically determining the fastest routes to the exits, based on the location of a fire and designated exit locations.  We use a fully distributed, self-organising method to generate evacuations routes, that also takes congestions in corridors during evacuation into account. The method is robust, fast, and fits well with readily available sensor network technology. The demonstration will illustrate the development of a wireless evacuation  system, and show its benefit over convential routing systems.

Smart UAV control
(Thales Research and Technology Netherlands)
SmartUAV is a mini UAV system developed by Thales Research and Technology Netherlands and TUDelft. It is designed to operate multiple UAV with a high degree of autonomy. In likely future scenarios, the UAV system is operating in a densely populated area and its safe operation is a primary concern.
However, the presence of unanticipated events in the real world requires human operator intervention. A novel design approach was used to design the human-machine-interaction aimed at making the UAV system, as a whole, more robust. The human operator is aided to see the relevant information in his work domain to help the automation of the system solve the control problems posed by the unanticipated event.
The human operators are an integral part of the system. The interface and automation design methods are derived from the field of Cognitive Systems Engineering (CSE). The result is a system that takes into account the human’s capabilities for information processing and control. In the event the automation is unable to resolve a problematic situation, the operator is supported to address the bottleneck at the level of control and the level of abstraction where it can be resolved. In other words, the operator does not need to fall back to a level of control or abstraction lower than that required by the situation, and the operator does not need to improvise on a higher level of control or abstratcion to circumvent limitations at lower levels. This approach is expected to result in accelerated development of automation and autonomy for these types of ‘intelligent’ platforms that operate in collaboration with humans in systems-of-systems. The approach was developed and applied during the design of SmartUAV. SmartUAV demonstrates how the design principles can be applied in concrete system.

SLAMMY: Second Life Answer Machine, a semantic network technology application
(TNO)
We have implemented a proof of concept Answer Machine allowing a user to ask a semantic network simple questions and receive responses, all in a natural multi-lingual fashion. The used technique is simple but shows the tremendous potential of semantic network technologies.
The kind of questions this method is be able to answer are simple ones like: “Who is the project-leader of ICIS?”, “Where is Ronald born?” and “What is the diameter of Mars?
Slammy is a front-end to the Answering Machine in Second Life, and gives an interface to the vast amount of knowledge contained in the underlying semantic network. Slammy is multilingual, and can converse in English, French, Dutch, German, Spanish and Italian, and demonstrates the versatily of semantic network-based knowledge representation.