Decentralized Collaborative Localization Based on Iterated Kalman Filter Using Relative and Absolute Observations

This paper proposes a robust algorithm which can realize distributed computing for the problem of multi-agent collaborative localization using relative and absolute observations. Firstly, the relative measurement model of agents is approximated by taking the state of their neighbors as prior knowledge, the approximation error can be modeled as the Gaussian distribution. This is very critical for the algorithm to achieve decentralized computing. Then the iterated kalman filtering algorithm is used to estimate the state for each agent using the information of itself and its neighbors. Finally, the proposed algorithm is compared with other existing approaches. Simulation results show that our algorithm provides better performance in positioning accuracy.