Simultaneous Self and Target Localization of Multi-Agent Systems with Noisy Relative Measurements Including Biases

In this study, we address a simultaneous self and target localization problem of multi-agent systems with noisy relative measurements, which is an important task to explore indoor space. First, we provide measurement models of the relative positions of agents and objects with sensor noise. The sensor noise contains bias as specific to each agent, which allowed for highly accurate localization. The relative posture is not measured in the setting. Next, we give a likelihood function of the unknown parameters: the positions and postures of agents, the positions of targets, and the biases. Then, we solve the corresponding optimization problem by the gradient-based dual ascent method on a Riemannian manifold, which yields a distributed algorithm to estimate all the unknown parameters simultaneously. The effectiveness of the proposed algorithm is verified through simulation.