Multi-Sensor Centralized Fusion for Unmanned Marine Vehicle Tracking With Constrained Communication

Real-time perception for unmanned marine vehicles (UMVs) accelerates the deployment of consumer unmanned electronic systems towards various autonomous marine applications. Compared to single-sensor configurations, multi-sensor fusion systems deliver significantly more accurate and comprehensive tracking results, and enable more efficient coordination of dispatched UMVs, promoting collaborative operations while minimizing resource waste. However, two manifest problems simultaneously challenge the realistic multi-sensor centralized fusion system: the origins of measurements acquired by sensors are ambiguous, and sensors transmit measurements to the fusion center (FC) usually with non-full-rate and also delay due to heavily constrained communication. These two problems eventually lead to the origin-unknown measurements arriving at the FC out-of-sequence with multi-step-lag. This paper first proposes the augmented state modeling to formulate the fusing of multiple multi-step-lag out-of-sequence measurements (OOSMs) with ambiguous origins in a unified framework, which not only provides filtered estimate for real-time tracking but also provides smoothed estimate for off-line analysis. Under this framework, the augmented state multi-sensor probabilistic data association (AS-MSPDA) algorithm is derived to effectively fuse multiple multi-step-lag OOSMs with ambiguous origins for improved target tracking in clutter. Numerical results demonstrate the proposed AS-MSPDA achieves the same optimal fusion accuracy as the state-of-the-art benchmark but with dramatically reduced overhead of computation and storage, meanwhile the smoothed estimate output by proposed AS-MSPDA is much more accurate than the filtered estimate, regarding as significant side-benefit for off-line analysis. Furthermore, numerical results also indicate the tracking benefits of fusing multiple multi-step-lag OOSMs tend to vanish as the communication constraint becomes very severe.