default Adaptive Bayesian behaviors for autonomous underwater vehicle surveillance networks


Adaptive Bayesian behaviors for autonomous underwater vehicle surveillance networks.  Goldhahn, Ryan A.; Braca, Paolo; LePage, Kevin D. CMRE-FR-2015-002. January 2015.

Autonomous underwater vehicles (AUVs) present a low-cost alternative or supplement to existing underwater surveillance networks. The NATO Centre for Maritime Research and Experimentation (CMRE) is developing multi-sensor data fusion techniques and collaborative autonomous behaviours to improve the performance of multistatic networks of AUVs for the purpose of underwater surveillance. In this work, a range-dependent acoustic model for the predicted probability of target detection is combined with the detections observed on all available platforms in a Bayesian framework to compute a posterior distribution on target position for each ping. This posterior is then used by the AUVs to collectively optimize their future actions based on a mission-driven measure of network performance. A Bernoulli filter is used to jointly estimate both the target state and whether zero or one target is present. Simulation results are presented quantifying the performance increase using these adaptive behaviours over traditional pre-planned trajectories.