MADDPG-Based Distributed Cooperative Search Strategy for Heterogeneous Agents System

The limited communication among agents is recognized as a significant constraint. It hinders and delays the collaborative exploration and exploitation of unknown environments. To tackle the challenge of cooperative search in communication-denied environments for agent swarms. We present a feature-based multi-agent reinforcement learning (MARL) framework. Firstly, we categorize agents into distinct roles based on their diverse characteristics and introduce a communication-complementary framework for multi-agent cooperation to maximize the benefits of individual agent characteristics. Secondly, we present a detailed introduction to the feature-based MADDPG algorithm, which effectively balances individual and collective benefits through a reward function. Finally, we assess the effectiveness of the proposed method through multiple simulations, showcasing its ability to effectively coordinate diverse agents.