Research Project SODA

Simulation-based Optimization using Differentiable Agents



     Source of funding:


Jun 01 2022 until May 31 2025

Dr. Philipp Andelfinger


German Research Foundation - DFG 


Design, development, and assessment of methods to enable efficient gradient-based optimization for a broad range of agent-based simulation models using automatic differentiation techniques. The methods are demonstrated and evaluated on optimization and calibration problems from the domains of road traffic and cell biology.

Agent-based simulations are used in the design and evaluation of systems in various domains. Frequently, optimization problems are solved by iteratively adjusting the parameters of the simulation model, which is associated with immense computation times. Gradient information, which could steer the optimization process towards local optima, cannot be applied to most agent-based simulations as explicit gradient expressions are typically unavailable.

The vision pursued in the project "Simulation-based Optimization using Differentiable Agents" (SODA) is to rely on automatic differentiation techniques to enable gradient-based optimization for a broad range of agent-based models. The key challenge is given by discrete model elements such as conditional branches, which are characteristic for agent-based models but can lead to unhelpful zero-valued or infinite partial derivatives. In a proof-of-concept publication, we showed that by manually substituting discrete model elements by smooth counterparts, the convergence speed and solution quality achieved in certain optimization problems could be vastly improved without substantial deviations from the original agent behavior. While this showed the significant promise of the approach on manually implemented example models, research is required to achieve a broad applicability and to systematically study the approach.
Hence, the project aims to answer the following research questions:

  • Can agent-based models be automatically translated to smoothed representations while retaining the per-agent behavior of the original models? As a naive smoothing renders even small simulations computationally intractable, knowledge of common agent-based model elements must be exploited.
  • To what degree can the model properties and computational structure of agent-based simulations be exploited to reduce overheads?
  • Within a given time and compute budget, is the approach able to identify higher-quality solutions than existing methods? High-dimensional optimization problems from transportation and cell biology serve as case studies to evaluate the approach.