Pressure Plate: Multi-Agent Requirement for Cooperative Learning
Abstract
The Pressure Plate project is an innovative approach to exploring multi-agent systems in cooperative environments. Multi-agent systems are pivotal in addressing problems requiring distributed intelligence, such as robotics, logistics, and gaming. This environment challenges agents to interact, cooperate, and strategize to achieve individual and collective goals. The primary aim of the project is to evaluate and improve learning algorithms in cooperative settings, emphasizing effective teamwork and interdependence among agents. The environment is structured to test the scalability and adaptability of learning strategies like Q-Learning and Deep Q-Networks (DQN). The Pressure Plate problem introduces a set of N agents, each assigned a unique goal in one of N levels. The task requires agents to navigate a grid-like environment with interconnected rooms containing pressure plates and doors.