Deep Reinforcement Learning (DRL) is a subfield of artificial intelligence that combines reinforcement learning and deep neural networks to solve complex problems. In multi-agent systems, intelligent agents interact simultaneously within an environment, and their decisions affect each other's behaviour. This paper examines Multi-Agent Reinforcement Learning (MARL), a significant branch of DRL, which is applied in systems with multiple agents having either common or conflicting goals. Key algorithms such as MADDPG, QMIX, and Mean Field RL, along with popular frameworks like TensorFlow, PyTorch, and Keras, are introduced. The applications of MARL in various domains, including economic systems, robotics, intelligent transportation, and resource management, are explored, and its advantages and disadvantages are discussed. Despite challenges such as high computational costs and limited scalability, MARL has the potential to drive significant innovations in technology and industry. The status of MARL in Iran and globally is analyzed, emphasizing the importance of investment and collaboration between academia and industry for advancement in this field. In conclusion, the paper highlights MARL's capability to solve complex problems and improve interactions, pointing to its potential in robotics, financial systems, and artificial intelligence.