Late Antiquity Model Projection - LAMP

This GSoC 2026 project proposes an advanced, open-source computational pipeline for the Late Antiquity Modelling Project (LAMP) to reconstruct historical movement and visual connectivity, moving beyond deterministic archaeological GIS with a data-driven, modular approach. The core objective replaces standard Least-Cost Path analysis with a Probabilistic Path Ensemble. Utilizing Maximum Entropy Inverse Reinforcement Learning (MaxEnt IRL), the pipeline autonomously learns historical cost weights by balancing terrain slope against surface friction derived from K-Means clustered SAR and Multispectral imagery. Additionally, the routing engine leverages spectral anomaly detection and structure tensors to uncover subtle pathways hidden in the landscape. The project's second pillar introduces 3D Viewshed Analysis. By processing high-resolution terrain geometry (.obj files) to calculate precise 3D lines of sight, the toolset enables researchers to evaluate the visual prominence and strategic interconnectivity of ancient settlements and outposts. Built over 12 weeks using robust Python geospatial libraries (Rasterio, GeoPandas, SciPy), the final deliverable is a scalable, mathematically rigorous toolkit. It will serve immediate LAMP needs at sites like El Bagawat while remaining seamlessly transferable to any archaeological context requiring advanced spatial modeling.