Robust Cooperative Localization Using Multi-Epoch Measurements Under Mismatched Measurement Models

High-precision and robust cooperative localization (CL) is essential for multi-agent systems. In practical applications, complex signal propagation environment and sensor faults can cause mismatches between the nominal and actual measurement models. This paper proposes a robust CL algorithm that accounts for the correlation of mismatched measurement models across multiple epochs. We use the state transition equation to constrain agent states over time and construct a likelihood function that integrates measurements from multiple epochs. Furthermore, we introduce high-dimensional latent variables to represent the unknown biases and noise characteristics caused by mismatched models and represent the uncertainty of these latent variables using the Normal-Wishart conjugate prior. Based on the Variational Message Passing framework, we derive closed-form solutions for the distributed inference of the posterior of the agent states and latent variables. The experimental results of multi-robot CL in indoor environments show that the proposed method has higher estimation accuracy and better robustness compared to the latest existing algorithms.