Surveillance in Anti-Submarine Warfare (ASW) has traditionally been carried out by means of submarines or frigates with towed arrays. These techniques are manpower intensive. Alternative approaches have recently been suggested using distributed stationary and mobile sensors, such as Autonomous Underwater Vehicles (AUVs). In contrast with the use of standard assets, these small, low-power and mobile devices have limited processing and wireless communication capabilities. However, when deployed in a spatially separated network, these sensors can form an intelligent network achieving high performance with significant features of scalability, robustness and reliability. The Distributed InFormation FUSION (DIFFUSION) strategy, in which the local information is shared among sensors, is one of the key aspects of this intelligent network.In this paper we propose two DIFFUSION schemes, in which the information shared among sensors consists of i) contacts, generated by the local detection stage; and ii) tracks generated by the local tracking stage. In the first DIFFUSION scheme contacts are combined at each nodes using the optimal Bayesian tracking based on the random finite set (RFS) formulation. In the second DIFFUSION scheme tracks are combined using the track-to-track association/fusion (T2T) procedure, then a sequential decision based on the association events is exploited. A full validation of the DIFFUSION schemes is conducted by the NATO Science and Technology Organization -Centre for Maritime Research and Experimentation during the sea trials Exercise Proud Manta 2012-2013 using real data. Performance metrics of DIFFUSION and of local tracking/detection strategies are also evaluated in terms of Time-on-Target (TOT) and False Alarm Rate (FAR).We demonstrate the benefit of using DIFFUSION against the local non-cooperative strategies. In particular DIFFUSION improves the level of TOT (FAR) with respect to the local tracking/detection strategies. Specifically, the TOT is increased over 90 − 95% while the FAR is reduced of two order of magnitude. The problem of communication failures, data not available from the collaborative AUV during certain periods of time, is also investigated. The robustness of DIFFUSION with respect to these communication failures is demonstrated, and the related performance results are reported here. Specifically, with 75% of communication failures the TOT is over 90 − 95% with a relatively small increase of the FAR with respect to the case of perfect communication.Index Terms-Collaborative data fusion, antisubmarine warfare, multistatic active sonar, target tracking, underwater sensor networks, autonomous underwater vehicles, real-world experimentation 1530-437X (c)

No abstract

Reverberation often limits the performance of active sonar systems. In particular, backscatter off of a rough ocean floor can obscure target returns and/or large bottom scatterers can be easily confused with water column targets of interest. Conventional active sonar detection involves constant false alarm rate (CFAR) normalization of the reverberation return which does not account for the frequency-selective fading caused by multipath propagation. This paper presents an alternative to conventional reverberation estimation motivated by striations observed in time-frequency analysis of active sonar data. A mathematical model for these reverberation striations is derived using waveguide invariant theory. This model is then used to motivate waveguide invariant reverberation estimation which involves averaging the time-frequency spectrum along these striations. An evaluation of this reverberation estimate using real Mediterranean data is given and its use in a generalized likelihood ratio test based CFAR detector is demonstrated. CFAR detection using waveguide invariant reverberation estimates is shown to outperform conventional cell-averaged and frequency-invariant CFAR detection methods in shallow water environments producing strong reverberation returns which exhibit the described striations.

No abstract

Many machine learning problems can be formulated as consensus optimization problems which can be solved efficiently via a cooperative multi-agent system. However, the agents in the system can be unreliable due to a variety of reasons: noise, faults and attacks. Providing erroneous updates leads the optimization process in a wrong direction, and degrades the performance of distributed machine learning algorithms. This paper considers the problem of decentralized learning using ADMM in the presence of unreliable agents. First, we rigorously analyze the effect of erroneous updates (in ADMM learning iterations) on the convergence behavior of multi-agent system. We show that the algorithm linearly converges to a neighborhood of the optimal solution under certain conditions and characterize the neighborhood size analytically. Next, we provide guidelines for network design to achieve a faster convergence. We also provide conditions on the erroneous updates for exact convergence to the optimal solution. Finally, to mitigate the influence of unreliable agents, we propose ROAD, a robust variant of ADMM, and show its resilience to unreliable agents with an exact convergence to the optimum.