Kinematic Contact Tracking Using Hybrid Features -- 20-022

The Navy envisions future multi-ship operations consisting of a diversity of sensors on manned and unmanned platforms against a peer competitor that involve multiple enemy surface and submerged targets. A capability is needed to track and identify these targets while operating in noisy, high clutter environments consisting of neutral vessels and natural clutter. This problem presents a number of unique challenges involving multi-sensor, multi-platform and multi-target tracking and identification. Traditional active sonar techniques for target tracking rely on highly uncertain kinematic measurements that often results in misassociations, large track uncertainty, multiplicity, purity and track-offs. These errors and uncertainties impact classification, leading to false alerts, missed true alerts and latent true alerts. The use of multiple sensors on multiple platforms offers the potential to dramatically improve target tracking and identification performance. However, new signal and information processing capability must be realized to achieve future Navy operational needs. Development in automation has recently shown significant progress. Nevertheless, currently deployed algorithms were designed for significantly simpler situations and only achieve reasonable performance for isolated sensor systems with few targets and low clutter. Todays’ approaches are inadequate to meet future operational requirements. Recent development on feature-aided tracking (FAT) as shown that including feature information into the trackers has the potential to improve association, tracking and classification. FAT is an extended tracker that uses features to mitigate incorrect data association, and to simultaneously conduct tracking and classification. FAT is ideally suited for CAS and is expected to provide a significant capability improvement for CAS and programmable PAS waveforms. The key take-away for active sonar FAT is that performance is critically dependent on the transmit waveform and on the selection of features. The coherent processing interval within a CAS repetition cycle is a free parameter that can be tuned depending on the local environmental conditions and operational scenario. The geometry and sonar ping-to-ping returns are relatively stable for high pulse repetition rate CAS processing; resulting in features that slowly evolve. FAT can leverage this information to improve correct measurement-to-track association and to mitigate incorrect measurement-to-track association, resulting in improved track solutions (i.e., purer, better continuity, fewer breaks, lower position error). We anticipate that FAT processing of CAS-like waveforms will improve tracking and classification performance in high clutter and low SNR environments. Combined with innovative signal processing for improved SNR and interpretable (physics-based) multi- spectral-aspect-scale features we anticipate achieving the objectives of this topic.