Multi-Agent Distributed Online Collaborative Target Search Based on Clustering Expectation Detection

In current distributed target search strategies utilizing information maps and receding horizon optimization, the finite prediction horizon may trap agents in local optima. We propose a distributed multi-agent online collaborative search method integrating clustering expected detection. Firstly, we design uncertainty and target probability maps using a two-dimensional grid map to characterize the search environment's uncertainty and target presence. Subsequently, we introduce an expected detection model, employing real-time K-means++ clustering of high-value regions to calculate expected detection areas. Finally, by incorporating expected detection gains into the utility function, we formulate an optimization objective to guide agents toward globally high-value regions. The method is tested in simulations and on real-world mobile ground robot platforms. Results illustrate its efficacy in guiding robots to escape local optima and achieve comprehensive environmental search. Compared with methods lacking expected detection, our approach accelerates target discovery within a limited timeframe. These findings offer insights into deploying collaborative search decision-making on mobile platforms in real-world environments.