default Performance achieved during REP 14 Atlantic and COLLAB/NGAS 14 trials


Performance achieved during REP 14 Atlantic and COLLAB/NGAS 14 trials. Sildam, Jüri; Canepa, Gaetano; Munafo, Andrea; Strode, Christopher; LePage, Kevin D.; Goldhahn, Ryan A.. CMRE-FR-2015-024. January 2016.

For the purpose of completeness of bi-static underwater system performance evaluation we propose a new diagnostic test of target perceivability based on known target locations and a set of contacts collected by the sonar system. Assuming exchangeability of target perceivability and target presence or state, our approach is based on a concept of channel capacity (C) defined in terms of maximum mutual information for a given target state distribution and the distribution of contacts defined by a signal-to-noise detection threshold (DT). In our experiments the DT value corresponds to the maximum C value (Cp) defined for an optimal bi-partition of contacts collected over a fixed number of sonar pings within a time window of O(10 minutes). The respective contact partition defines the probability of target detection (Pd) and the probability of false alarms (Pfa) achievable under Cp. The time series of Cp, Pd, and Pfa were obtained by sliding the time window along the target trajectory. We demonstrate a successful test by applying the algorithm to data collected by underwater autonomous vehicle systems operated during two trials in 2014. Target tracks and visually observable target detections match well the observed Pd and Cp peaks.