Unscented Kalman Filtering for Localization using Range or Bearing Data

A search and rescue localization algorithm is proposed using an unscented Kalman filter (UKF). The UKF provides superior performance to other filtering options such as the extended Kalman filter or the particle filter and is computationally less expensive than neural network localization methods. Each search agent obtains asynchronous measurements of its range and/or bearing relative to the victim. It is assumed that the victim's pose (position and orientation) and speed estimates and the associated covariance matrices are available to all agents simultaneously. Each agent maintains its own UKF to update these estimates and its pose estimate when a new measurement is received. Simulation results demonstrate that the agents can track the victim's location and orientation with minimal steady-state error despite significant measurement noise and large initiate estimation error.