Inter-level relations between models in ATM

From ComplexWorldWiki
Jump to: navigation, search

Contents

ComplexWorld PhD

Nataliya Mogles, VU University Amsterdam

Supervisor: Jan Treur

Daily supervisor: Tibor Bosse

Overview

According to the SESAR programme, the future ATM system will consist of a large number of (both human and automated) agents at different levels of the ATM process, which collaborate in a sophisticated and resilient manner order to achieve an optimal performance with minimal chance for hazardous events. To study the emergent behaviour of this complex system, the current project proposes to exploit computational modeling techniques. More specifically, the proposal is to model the behaviour of the ATM system along three abstraction or aggregation dimensions (the temporal dimension, the process abstraction dimension, and the agent clustering dimension)[1]. For each dimension, computational models will be developed at different levels of aggregation. By establishing formal (mathematical and logical) relations between the models at the different levels, it will then be possible to relate emergent phenomena to (and explain them in terms of) local mechanisms. Thus, the developed computational models and interlevel relations will provide more insight in the types of skills and properties that are required for (both human and software) agents involved in ATM processes, to ensure emergence of optimal performance of the overall system with minimal errors. Such insights may be used, among others, to increase awareness of weaknesses and bottlenecks in the organization, and to develop more effective training methods for human operators as well as more effective automated systems.

The main research questions of this project are:

What is resilience of a system and what makes a system resilient?

What is complexity and emergence of complex behaviour of a system?

What kind of classification dimensions, or levels, can be of interest if we consider interlevel relations, both in ATM and in other domains?

What types of modelling methodologies exist in ATM?

What type of modelling methodologies exist in other domains and could be useful for ATM?


This project is closely related to such chapters of the ComplexWorld position paper as Emergence and Resilience (see ComplexWorld Position Paper).

Context

To study the behavior of such a complex system as air traffic management in a detailed manner, approaches are needed that are able to deal with this complexity. Typical questions that need to be addressed are: How can descriptions at a global level of the system be related to descriptions at local levels and the organization of interactions? How does a change in the behavior of a local component or of the interaction organization impact the behavior of the global system? Can descriptions be found of the behavior at the global level that approximate the behavior of the local elements combined, but abstract from the local details? Existing (mainly mathematical) approaches to study the behavior of the ATM system (often focused on safety analysis) do not always make an explicit distinction between these levels of aggregation, or only concentrate on one specific level. For example, STAMP[2] focuses on the aggregated dynamics of the ATM organization as a whole, whereas TOPAZ[3] takes a local perspective, focusing on the behavior of individual actors. Although both types of approaches have proven their usefulness, separately, none of them provides much insight in how the local mechanisms and their organization cause the global behavior.

Objectives

The current project proposes to address this question from a computational modeling perspective, by defining explicit interlevel relations between computational models at different levels(see [4]). The underlying assumption is that the behavior of the ATM system can be described at different levels of abstraction or aggregation. These levels may concern different dimensions: (1) the temporal dimension, from transitions over small time steps to patterns over longer time durations, (2) the process abstraction dimension, from physiological functioning of agents via cognitive and affective processes to behaviors, and (3) the clustering dimension, from individuals to subgroups to teams as a whole; see also The approach that is proposed to perform such an analysis is to use computational modeling to formalize the behaviors at the different levels of abstraction of the ATM process, and one step further, to relate the formal descriptions at the different levels to each other via computational means as well. By explicitly developing these computational models and their mutual interlevel relations, the modeler will gain more insight in the properties that are required for the agents at the local levels to ensure the behavior at the global levels. More concretely, the models will enable the modeler to identify the types of skills and characteristics that are required for (both human and software) agents involved in ATM processes, and the organization of their interaction to ensure optimal performance with minimal errors. Such insights may be used, among others, to develop more effective training methods for human operators as well as more effective automated systems, and suggestions concerning weaknesses in the organization and how to overcome these.


Work Done

  • Formalisation and agent-based analysis of a real world aviation incident at local and global levels of process abstraction and time dimensions[5]
  • Agent-based dynamic model, simulation and analysis of an aviation incident at local and global levels of process abstraction and time dimensions[6], [7].
  • Comparison of four popular approaches of accident modeling and analysis in aviation: agent-based modeling, FRAM, STAMP and Event Trees[8]
  • PhD thesis with the title "Agent-Based Analysis and Support of Human Functioning in Complex Socio-Technical Systems: Applications in Safety and Healthcare" completed and defended on the 8th of April, 2014

PhD Thesis - Short Summary

In the thesis entitled "Agent-based analysis and support of human functioning in complex socio-technical systems: Applications in Safety and Healthcare" methods for agent-based analysis and support of human functioning in complex socio-technical systems have been addressed. The methods are in accordance with a three-dimensional framework for classification of agent-based models, and enable the modeler to develop models at different abstraction levels, and to establish interlevel relations between those models. Models of human functioning were created and analyzed in safety and healthcare domains with the aim to provide support and recommendations both at the level of the whole system and at an individual level. More specifically, Part II of the thesis is dedicated to the analysis of multiple components of a safety critical socio-technical system and Part III is focused on the analysis and support of humans in socio-technical systems related to the healthcare domain.


References

  1. Bosse, T., Hoogendoorn, M., Klein, M. C., & Treur, J. (2010). A three-dimentional abstraction framework to compare muti-agent system models. Computational Collective Intelligence: Technologies and Applications, Proceedings of the Second International Conference on Computational Collective Intelligence. 6421, pp. 306-319. Springer Verlag.
  2. Leveson, N. (2004). A new accident model for engineering safer systems. Safety Science, 42, 237-270.
  3. Blom, H.A.P., Bakker, G.J., Blanker, P.J.G., Daams, J., Everdij, M.H.C., and Klompstra, M.B. (2001). Accident risk assessment for advanced air traffic management. In: Donohue, G.L., and Zellweger, A.G. (eds.), Air Transport Systems Engineering, AIAA, pp. 463-480.
  4. Bosse, T., Hoogendoorn, M., Klein, M. C., & Treur, J. (2010). A three-dimentional abstraction framework to compare muti-agent system models. Computational Collective Intelligence: Technologies and Applications, Proceedings of the Second International Conference on Computational Collective Intelligence. 6421, pp. 306-319. Springer Verlag.
  5. Bosse, T., and Mogles, N.M. (2012). Formal analysis of Aviation Incidents. In: Proceedings of the Twenty-Fifth International Conference on Industrial, Engineering and Other Applications of Applied Intelligent Systems, IEA/AIE’12.
  6. Bosse, T. and Mogles, N. M. (2013). Studying Aviation Incidents by Agent-Based Simulation and Analysis. In: Proceedings of the Fifth International Conference on Agents and Artificial Intelligence, ICAART’13
  7. Bosse, T., & Mogles, N. M. (2014). An Agent-Based Approach for Accident Analysis in Safety Critical Domains: A Case Study on a Runway Incursion Incident. In Transactions on Computational Collective Intelligence XVII (pp. 66-88). Springer Berlin Heidelberg
  8. Bosse, T. and Mogles, N. M. (2013). Comparing Modelling Approaches in Aviation Safety. In: Proceedings of the Third International Air Transport and Operations Symposium (ATOS), ATOS’13
Personal tools
Namespaces

Variants
Actions
Navigation
Toolbox
Share