Sign in

Minds for Mobile Agents


The research questions, and scientific and practical objectives motivating this project emerged from the Data vs. Corona initiative. This initiative aimed to improve physical distancing by simulating, explaining, and predicting the movement trajectories of individuals and groups of pedestrians performing complex tasks involving a series of spatially defined goals in low to medium density settings (e.g., grocery shopping, visiting restaurants). The mathematical “Predictive Pedestrian” model developed to address these questions allows the incorporation of complex series of goals and individual differences in the way in which pedestrians move around and interact with each other (see here for background and illustrations), but is computationally expensive. The M4MA package provides an efficient computational solution that makes it possible to realize the promise of the Predictive Pedestrian model to quantitatively characterize and simulate pedestrian behaviour in a wide range of scenarios. It did so by profiling the the Predictive Pedestrian model implemented in a collection of R language functions to identify computational bottlenecks. A selection of these functions was then ported to an R package and options added to perform the associated computations in C++ through the Rcpp framework, providing a speed up by approximately one order of magnitude.

Participating organisations

Environment & Sustainability
Netherlands eScience Center
University of Amsterdam



Dora Matzke
Principal investigator
University of Amsterdam
Eva Viviani
Research Software Engineer
Netherlands eScience Center
Malte Lüken
Malte Lüken
Research Software Engineer
Netherlands eScience Center
Charlotte Tanis
Tessa Blanken
Michael Lees
University of Amsterdam
Niels  Drost
Niels Drost
Programme Manager
Netherlands eScience Center

Related tools



An R package containing C++ implementations to speed up the simulation and parameter estimation of the Predictive Pedestrian model.

Updated 4 weeks ago