Intelligent Selection of Waveform Based on Predicted Target State for Active Sonar

Abstract

Underwater target tracking has vital importance in a variety of applications, most notably military surveillance and defense. This work is focused on underwater target tracking using active sonar. In active sonar, sound pulses are transmitted in water, and an array of hydrophones receives echoes produced by the target and any other obstacles. The received echoes are processed to obtain estimates of object location and radial velocity, which form the input to a tracking filter for estimating the object's motion over time. The object location is given in terms of range and bearing (angle); the object's radial velocity, also known as range-rate, is obtained from the estimated Doppler shift. In active sonar, the conventional sound pulses are the single-frequency continuous wave (CW) pulse and the frequency modulated (FM) pulse. These pulses differ from one another in terms of the range resolution they provide and whether or not they allow the system to measure Doppler shift. CW pulses allow for measurement of Doppler shift but have relatively poor range resolution. In contrast, a broadband FM pulse is insensitive to Doppler but has a better range resolution than a CW pulse. Reverberation is defined as the rejection of sound energy toward the sonar system by objects other than the target of interest. Most active sonars operate in a reverberation- limited environment where an appropriate choice of sound pulse plays a key role in target tracking. If the target exhibits high Doppler at the time of sound pulse impingement, a CW pulse provides better tracking capability, while an FM pulse is more effective for a low Doppler target. A challenge, however, is that the target Doppler changes with time and is unknown at the time of waveform impingement. At present, the complementary strengths of CW and FM waveforms can be exploited only by an operator who actively chooses between them. For the operator, to choose the transmission waveform for each ping presents an unreasonable burden. In this thesis, we devise and evaluate a decision algorithm that decides which waveform to transmit based on the target's predicted state. Building on the intuition of selecting a waveform based on the target's Doppler, we propose the Predicted State-Based Selection (PSBS) algorithm, which uses an estimate of the target's Doppler, derived from a state prediction produced by the tracking filter, to select the waveform for the next ping. A sys- tem consisting of a target path simulator, measurement data simulator, and tracking filter is modeled to conduct Monte Carlo simulations of PSBS. Its performance is evaluated by comparing it to the performance of transmitting either CW always or FM always. Simula- tion results show that the PSBS algorithm improves target localization estimates by 7.7% on average from the next best performing waveform selection approach, over all target turn rates considered. While PSBS improves significantly upon using only CW or only FM waveforms, it suffers substantial performance degradation for highly maneuvering targets. To address this shortcoming, this thesis suggests an enhancement to the PSBS approach: EPSBS (Enhanced PSBS) takes into account the number of measurements being considered by the tracker when making a waveform decision. Simulation results show that EPSBS eliminates the performance degradation observed for PSBS when targets undergo rapid maneuvers. Based on results obtained in this thesis, we conclude that intelligent selection of transmitted waveforms` can significantly improve tracking performance in active sonar.