Anatolia contains the remains of cities connected by a network of roads, which reveals much about the culture of the past. Previous research on the caravanserais on the ancient Silk Roads during the Anatolian–Seljuk period lacks a global vision mapping historical phenomena with precise spatial coordinates. Although there has been a substantial amount of research on the period, it consists of individual projects shedding light on specific aspects, rather than on the overall network of buildings. In contrast, this current work investigates whether entire trade networks can be simulated employing artificial intelligence to unravel the optimal paths in between the caravanserais distributed on a three-dimensional digital topography. Referring to the spatial distribution of the caravanserais located in a “day’s travel,” we aim to geolocate these historic buildings of Anatolia. The methodology consists of three main phases: surveying existing publications on caravanserais, manually locating well-documented remaining historical buildings in the geographic information system, and estimating their location through supervised learning. The dataset of images from caravanserais allows us to train a convolutional neural network to locate the missing caravanserais. An accurate database allowed the construction of a digital topography in the Unity game engine, in which an agent discovers the optimal paths using deep reinforcement learning (DRL). Based on the simulations, the research aims to uncover the ancient trade networks in Anatolia.
The Routledge Companion to Artificial Intelligence in Architecture