Removing Ping Timing Ambiguity via Data Association Vishal C. Ravindra, Marco Guerriero, Peter Willett and Shengli Zhou Department of Electrical and Computer Engineering University of Connecticut Storrs, CT 06269 Stefano Coraluppi NATO Undersea Research Centre Viale S. Bartolomeo 400, 19126 La Spezia, Italy Abstract— This paper considers the use of a multiple ping active sonar approach in order to track multiple targets. In most underwater target tracking applications that rely on active sonar observations, the sonar sends out a single ping at each scan interval and receives returns that might either be from target(s) or due to clutter. We propose to use a multi-ping paradigm, with the idea that the better detectability of more pings can, assuming a high clutter density, lead to better localization of targets. However, this introduces a timing ambiguity in the ping returns adding a new level of complexity in the data association. In this paper we propose a Multi-Ping Data Association (MPDA) algorithm as a solution. MPDA formulates the assignment problem as a linear program, which could then be solved using a primal-dual interior point approach. A comparison is made between three different scenarios: (a) the sonar sends out a single ping in each scan interval; there is no timing ambiguity (which return is due to which ping), (b) the sonar sends out multiple pings within a scan interval and the timing ambiguity is avoided by using orthogonal waveforms for each ping, (c) when the sonar sends out multiple pings, each ping being an identical waveform, within each scan, leading to a timing ambiguity. A well known multiple target tracking technique such as the JPDA is used in cases (a) and (b), while the MPDA algorithm is used to solve the assignment problem in case (c), and is shown to perform better than case (a) in high clutter densities. Case (b), the unrealizable bogey, performs best. I. I NTRODUCTION Multiple-target tracking (MTT) is an essential requirement of surveillance systems which consist of sensors, communi- cation links, as well as computer subsystems to interpret the measurements of the sensors [1], [2]. The sensors, such as radar, infrared, sonar, etc. report measurements from diverse sources including targets, background noises (e.g. sea clutter), etc. Since there are multiple sensors observing multiple targets, there is a need for data association, i.e., the association of measurement data to targets. The accuracy of data association is critically important since any misassociation of data to wrong targets or to clutter will often lead to tracking fail- ures such as the cross-over of tracks generated by parallel targets, or loss of tracks, as well as false track initiations. In underwater scenarios especially, a high clutter and a low signal-to-noise ratio (SNR) environment (i.e., for very low observable targets) present many challenges making the as- sociation task even more difficult. There have been many approaches in the literature to solve the problem of data association for target tracking. One of the earliest methods which deals with measurements obtained from only one scan at a time, is the Global Nearest Neighbor (GNN) method [3], which formulates the measurement-to-track association as a 2-dimensional assignment problem and chooses among all possible assignments of measurements to tracks, choosing the assignment with the highest probability for tracking. This method is computationally intensive even after applying data preprocessing techniques such as gating. JPDA [5] is a target- oriented Bayesian approach which is a recursive technique. The JPDA does not make “hard” associations like the GNN algorithm, instead it associates an established track with all the measurements falling inside its validated gate within the scan period. These “soft” associations of a track with each measurement are nothing but measurement-to-track posterior probabilities. The JPDA performs well in clutter, since the fused measurement associated to a track is the combination of the weighted probabilities of all the measurements in the validation region. Since the weights are nothing but posterior probabilities, the fused measurement typically includes a sig- nificant contribution from the target originated measurement while marginalizing the clutter originated measurements. In underwater multi target tracking applications, two general types of sonars are used: passive and active. Passive sonar systems are used primarily to detect noise from marine objects, such as submarines, ships, and marine animals like whales. Unlike the active sonar, a passive sonar does not emit its own signal, which constitutes an advantage for military vessels that do not want to be detected or for scientific missions that concentrate on listening to the ocean. Rather, it only detects sound waves coming towards it. Thus, a passive sonar cannot measure the range of an object unless it is used in conjunction with other passive listening devices. Multiple passive sonar devices may allow for triangulation of the sound source. Active sonar transducers emit an acoustic signal or ping of sound into the water. If an object is in the path of the sound pulse, the sound bounces off the object and returns an ”echo” to the sonar transducer and a “contact” is termed to have been established. If the transducer is equipped with the