Differential Evolution (DE) is a population-based stochastic optimization algorithm that iteratively refines candidate solutions through mutation, crossover, and selection operations applied to real-valued parameter vectors, making it well-suited for calibrating complex, non-linear models such as social force models for pedestrian dynamics. In the context of crowd modeling and pedestrian simulation, DE is particularly valuable because it enables efficient parameter estimation for models whose objective functions are non-differentiable or highly multimodal, circumventing the limitations of gradient-based methods. Key variants include adaptive schemes that dynamically adjust mutation and crossover rates, as well as hybrid formulations that combine DE with local search strategies to improve convergence speed and solution quality when fitting model parameters to observed crowd trajectory data.
Dictionary term