This project aims to explore the feasibility to create a branch of Regional Science that takes its starting point in complexity science, and complexity economics. The key is to leverage large scale simulations of agents (individuals, households, firms etc) to form a better understanding of the dynamics of spatial and temporal behaviour in such settings as housing choice, activity patterns, energy use and labour marked decisions. The key questions that we aim to study are the computational challenges that agent interactions pose, and methods to calibrate or estimate the parameters of such simulation systems. Agent based simulations of markets or travel patterns tend to be difficult to parallelize due to the need to synchronize when agents interact, for instance on a congested road. We will experiment with different trade-offs between verisimilitude and computational tractability. Estimating the parameters of a simulation model with a large number of interacting agents is not yet a solved problem. In this project we will experiment with looking at novel approaches that combine multiple coarse grainings, inspired by both bayesian methods and generative machine learning techniques. This project is an initial exploration of the feasibility of some ideas on how to implement simulation software using Chapel in order to get the full benefit of a compute cluster. We will pursue funding from e.g. Riksbankens Jubileumsfond and VR. Meanwhile it is useful to explore the feasibility of the intended toolchain.