Objectives To address the problems of insufficient adaptability of region-partitioning strategies under node failures and the tendency of probability-driven search methods to cause redundant coverage and local clustering in cooperative search of underwater moving targets by surface robot swarms, a cooperative receding-horizon search strategy is designed to account for both target probability cues and balanced coverage.
Methods A target existence probability map and a coverage information model are constructed. The motion uncertainty of the underwater moving target is represented by probability diffusion, and the target existence probability is updated by integrating cumulative detection probability (CDP) decisions with Bayesian inference. Within a receding-horizon optimization framework, a coverage reward, future trajectory exchange, and field-of-view overlap penalty are introduced; model predictive control and particle swarm optimization are then employed to solve for the robots' motion control sequences.
Results Monte Carlo simulations with 500 trials demonstrate that the proposed method achieves a target discovery probability of 79.8%, with an average discovery time of 34.98 min for successful trials. Compared with the multi-surface-robot partition-based cooperative search method, the proposed method exhibits a trend toward a higher target discovery probability and reduces the average discovery time of successful trials by 4.32%. Under single-robot failure conditions, the average target discovery probability across the four node failure scenarios increases by 5.4 percentage points, and the average discovery time for successful trials decreases by 3.49% overall, although a slight increase occurs under an individual node-failure scenario. Ablation studies show that removing the coverage reward term decreases the target discovery probability by 5.4 percentage points, while the future overlap penalty helps reduce conflicts among the planned search regions of multiple robots.
Conclusions The proposed method enables soft coordination among multiple robots without fixed partition constraints, reduces redundant searching, and improves overall search performance and robustness to single-robot failures.
DownLoad: