Optimization and Control of Agent-Based Models in Biology: A Perspective

Area of Science:

• Computational Biology and Life Sciences
• Complex Systems Modeling

Background:

• Agent-based models (ABMs) are crucial for life sciences, especially for poorly understood systems where equation-based models fail.
• Analyzing and controlling large-scale ABMs is challenging due to limited mathematical tools, relying primarily on simulation.
• Existing mathematical frameworks for ABM analysis are insufficient for complex optimization and control tasks.

Purpose of the Study:

• To propose a novel mathematical framework for optimization and control of agent-based models (ABMs).
• To treat ABMs as surrogates of actual systems, enabling the application of established mathematical tools.
• To outline a research program for developing and applying these new mathematical techniques to ABMs.

Main Methods:

• Viewing ABMs as surrogates for real systems, rather than models themselves.
• Modeling the surrogate system using data from the ABM within frameworks like differential equations or difference equations.
• Applying optimization and control strategies to the surrogate model and lifting solutions to the actual system.

Main Results:

• Demonstrated the approach using Sugarscape and a consumer-resource ABM, illustrating dimension reduction and approximation techniques.
• Highlighted the significant mathematical challenges in applying this method to large, complex ABMs.
• Showcased the potential of using established mathematical tools for ABM optimization and control.

Conclusions:

• The proposed surrogate modeling approach offers a promising avenue for enhancing the mathematical analysis of ABMs.
• Significant mathematical research is required to overcome the challenges for applying this to complex, large-scale ABMs.
• This work lays the foundation for a new research program in ABM optimization and control within the life sciences.