<p>This paper investigates cooperative target estimation for multiple strapdown vehicles under correlated noise and random observation delays. A distributed fusion estimation algorithm is proposed that explicitly accounts for stochastic delay characteristics and multi-rate sampling. First, a random observation delay model is developed based on noise correlation, and an equivalent delay-free system is derived through state equation transformation. Then, local optimal filters are designed for each strapdown subsystem using the projection theorem, including predictions for unsampled time points. Cross-covariances between subsystems are computed to enable matrix-weighted fusion of local estimates. To further enhance robustness and estimation accuracy, a feedback-based distributed fusion strategy is introduced by minimizing the trace of the global error covariance matrix. Simulation results under multiple scenarios demonstrate that the proposed approach improves both estimation accuracy and robustness in complex environments, offering theoretical support and methodological reference for practical cooperative sensing applications.</p>