Recursive Bayesian Initialization of Localization Based on Ranging and Dead Reckoning
This work addresses a specific challenge in cooperative localization for robotics or autonomous systems, but it appears incremental as it builds on existing particle filter techniques.
The paper tackles the problem of initializing state estimation in localization using ranging and dead reckoning, proposing a particle filter method to achieve a uni-modal estimate with low covariance, with results illustrated through simulations and experimental data.
The initialization of the state estimation in a localization scenario based on ranging and dead reckoning is studied. Specifically, we start with a cooperative localization setup and consider the problem of recursively arriving at a uni-modal state estimate with sufficiently low covariance such that covariance based filters can be used to estimate an agent's state subsequently. A number of simplifications/assumptions are made such that the estimation problem can be seen as that of estimating the initial agent state given a deterministic surrounding and dead reckoning. This problem is solved by means of a particle filter and it is described how continual states and covariance estimates are derived from the solution. Finally, simulations are used to illustrate the characteristics of the method and experimental data are briefly presented.