Acoustic Source Localization And Tracking With Markovian State-space Models

Passive underwater acoustic source localization and tracking has been of great inter-est for both research community and various application users. Traditional focus is on the detection of target with low-level noise-like self-radiation. The large-aperture receiving ar-ray and complex processing systems are used in these applications. Lately, the localization problem of those moving platforms with a cooperative acoustic source is more widely dis-cussed, as different types of autonomous underwater vehicles become popular in a variety of applications. Most of the localization methods currently used is evolved from the radio technology. Those methods rarely consider the impact introduced by the underwater acous-tic channel. They could work well in deep sea, but the performance will degrade in shallow water because of the waveguide effect. In shallow water environments, fluctuations along the boundaries and in the media will cause severe multi-path time-delay spreading and Doppler spreading, which can be more serious for a moving platform. The shallow water environment is a doubly-spread and rapidly time-varying channel for sound propagation. This thesis is aimed to track the moving platform carrying with a cooperative acoustic source via a small receiving array in shallow water.When tracking a cooperative target, the array received signals from source are d-ifferent from both the active localization and passive localization. The receiver knows the source signals’ parameters for cooperative source tracking. The localization is sim-plified from three-dimension to two-dimension as the Autonomous Underwater Vehicle often move at a constant depth. With the knowledge that the source state always changes continuously as a wide-sense Markov process, we can model the source movement in a kinematic equation. Then the tracking process is the same as random process estimation with some measurements. For the moment, it will suffice to generalize the tracking prob-lem to the state-space model, Given the tracking problem described by the state-space model, we wish to operate on the observable inputs and observable outputs to find the min- imum mean-square-error estimators of the source state which is not directly observable. However, instead of the unavailable inputs, we drive the model with a measurement error. The measurement vector consists of source bearing and range difference, both of which are estimated via a five-element uniform linear array developed in this thesis. The source bearing angle is estimated via beamforming and the range difference is calculated with the time difference of transmitting-receiving interval of signals when the transmitting interval is given. Sequential Bayesian filter is used to realize the tracking when system is described by the state-space model.Because of the waveguide effect for signal propagation in the shallow water environ-ment, the estimation error will be large if traditional estimation methods such as cross-correlation are used to estimate the time interval of receiving pulse signals. The estima-tion accuracy could be improved after time-reversal processing for receiving signal. Time-reversal processing is a signal processing method developed based on the reciprocity of sound propagation in a time-invariant environment. When the environment is stationary or changing slowly, time-delay spreading caused by multi-path propagation can be com-pressed with time-reversal processing. Passive time-reversal is adopted to recover the re-ceiving signal used to estimate time interval of the receiving pulse signals. The performance of the processing is verified. However, in the fluctuating shallow water environment, the performance of the time-reversal method will be limited. For cooperative source tracking problem we focus on in the thesis, the waveform optimization can be adopted to improve the performance of the tracking method.The signals of different paths are encountered with different time-doppler spreading when they propagate in fluctuating time-varying shallow water. The time-doppler ambi-guity function of the receiving signal have an extention around several regions. The time-doppler spread introduced by relative movement between source and receiver is stable or changing slowly compared with that introduced by surface waves and/or internal turbu-lence. Based on the knowledge of acoustic transmitting character of echolocating mammal-s such as bats and dolphins, we adopt the signal with good resolution both in time-delay and doppler scale. Time-doppler spread of the direct arrival signal or first arrival signal is tracked after the computation of time-doppler spread function. Then it is used to calcu-late the measurement vector of the state-space model for tracking the moving source in fluctuating time-varying shallow water.When moving in some particular trajectory, the source can not be tracked by a station-ary observer. The observer need to move with some strategy to keep the observability of the source. The optimal observer moving path is searched in the mean-square-error sense. The source moving in a particular trajectory is observable after the observer’s movement.Localization and tracking of a cooperative moving source in waveguide is verified via simulations in this thesis. The performance of combination of time reversal processing and the tracking method is verified in a time-invariant waveguide. The performance of combination of delay-doppler filtering and the tracking method is verified in a time variant waveguide. For source moving in a particular trajectory, the trajectory of the receiving array is searched based on mean square error in order to improve the observability of the moving source. During the thesis research, a prototype localization system has been designed and built, including a simulated source and a passive sonar system. Through a tank test and sea experiment, the system has shown working reliably while acquiring a mount of valuable real data. The processing results of the experiment data has verified the performance of a combination of time reversal processing and tracking proposed in the thesis.