Multi-Agent Reinforcement Learning Framework

The Multi-Agent Reinforcement Learning Framework builds a virtual copy of a real foodshed and simulates how climate change drives migration, food security and rural population change. Thousands of agents — households, cooperatives, regional managers and policymakers — make decisions season by season and learn how to adapt, driven by real climate and Earth-observation data. Through a simple chat interface, users describe a scenario in plain language and receive interactive maps, timelines and plain-language reports with policy recommendations. It lets decision-makers test the likely impact of adaptation policies before committing to them in the real world.