NIMBioS Investigative Workshop
Optimal Control and Optimization for Individual-based and Agent-based Models
Group photo of NIMBioS Investigative Workshop on Optimal Control and Optimization for Agent-based Models
Topic: Optimal Control and Optimization for Individual-based and Agent-based Models
Meeting dates: December 1-3, 2009.
Organizers:
Filippo Castiglione (Institute for Computing Applications, Rome);
Volker Grimm (UFZ Center for Environmental Research, Leipzig);
Reinhard Laubenbacher (Virginia Bioinformatics Institute);
Suzanne Lenhart (Univ. of Tennessee, Knoxville)
| Agenda (PDF) | Participants | Evaluation report (PDF) |
Objectives: Agent-based models are used increasingly to understand a broad range of biological phenomena, including, e.g., tumor growth, the immune system, and the spread of infectious diseases across social networks. In all these cases it would be very useful to have analytic methods available to study in general how possible interventions would affect system dynamics. The advantage of agent-based models is that they integrate local relationships to capture global emergent dynamics, without needing global parameters as input. The disadvantage of this type of model is that very few mathematical analysis methods are available to produce general descriptions of model response particularly in terms of spatio-temporal patterns arising from even fairly simple ABMs. In particular, the absence of a state space description of ABMs makes it very difficult to apply available control theory methods to study effective interventions. Applications of ABMs in situations with possible interventions by human actions (e.g. vaccination and quarantine schemes) have usually been limited to scenario analyses. In this case the models are simulated numerous times to compare alternative scenarios for intervention.
One possible approach to this problem is to construct state space models that approximate the agent-based model, similar to approaches proposed for discrete event simulations. This uses system identification methods developed for the state space model framework for agent-based simulations. Control-theoretic approaches for this modeling framework have been explored in a few cases. A first exploratory project in this direction resulted in a control method for in vitro competition of viruses. Such methods from approximate models may not work when there is spatial heterogeneity in the agent-based model (Federico, Gross and Lenhart, in preparation). Various techniques from optimal control and discrete optimization should be considered to investigate alternative formulations of control in relation to a state-space approximation and then compared to a similar formulation applied to the ABM.
This workshop brought together researchers working in agent-based models, optimal control and optimization to discuss the possible development of control theoretic approaches for agent-based models, beginning with the ones mentioned above. Alternative formulations of the approximation models and optimal control/optimization methods appropriate to each formulation were considered.
Summary Report. The workshop began with an overview of some optimal control techniques. Among the other presentation topics were strategies involved in constructing IBMs, protocol for standardizing the reporting of results, epidemic IBMs with adaptive behavior, emergence behavior from ABMs, and the use of ABMs in films and games. Discussion sessions highlighted the fact that there are a variety of different types of ABMs and IBMs. Other topics of interest included algebraic-based control methods; aggregate model approaches to aid in 'optimal' control of IBMs; and high performance computing and large simulations in relation to IBMs.
Participant Podcasts
Here's what participants are saying about their experience at the NIMBioS Optimal Control Investigative Workshop.
Ben G. Fitzpatrick, Loyola Marymount University, Clarence J. Wallen, S. J. Professor of Mathematics
Katarzyna Rejniak, Faculty Member, Integrative Mathematical Oncology, H. Lee Moffitt Cancer Center & Research Institute
Volker Grimm, Senior Scientist, Helmholtz Center for Environmental Research
NIMBioS Investigative Workshops focus on broad topics or a set of related topics, summarizing/synthesizing the state of the art and identifying future directions. Organizers and key invited researchers make up half the 30-40 participants in a workshop, and the remaining 15-20 participants are filled through open application from the scientific community. Open applicants selected to attend are notified by NIMBioS within two weeks of the application deadline. Investigative Workshops have the potential for leading to one or more future Working Groups. Individuals with a strong interest in the topic, including post-docs and graduate students, are encouraged to apply. If needed, NIMBioS can provide support (travel, meals, lodging) for Workshop attendees.
