Roomenv V0

RoomEnv-v0 is a challenging environment compatible with Gymnasium, designed to test and improve an agent's performance in partially observable Markov decision processes (POMDPs). This environment simulates a large room where multiple characters can move freely and place objects. The agent earns rewards by observing the characters' actions and answering questions about the locations of objects. At its core, it leverages structured RDF triple data, combined with common-sense knowledge graphs such as ConceptNet, to enhance decision-making accuracy. The application value of this environment lies in researching and developing machine learning models that exhibit characteristics of human memory systems, particularly in knowledge reasoning and memory management capabilities within complex and dynamic scenarios.