Accurate navigation of autonomous underwater vehicles (AUVs) is a key task for data collection at sea with high resolution in time and space. Sidescan sonar (SSS), originally developed for imaging the seafloor, has high potential for establishing landmark-aided navigation or simultaneous localization and mapping (SLAM) capabilities on small-scale AUVs. A key task to establish these capabilities is to determine the height of the AUV from the seafloor, also referred to as “flight height.” This paper combines image processing techniques with probabilistic data association to detect and filter AUV flight height from SSS data. The proposed method first aims to detect the edge between the water column and the seabed using image processing techniques. (The pixel index of this edge is proportional to the flight height in meters.) Subsequently, a multisensor probabilistic data association filter (PDAF) fuses the resulting flight height detections computed from images provided by port and star-board SSS transducers to effectively mitigate missed detections and false positives. To facilitate deployments, we evaluate the performance of the proposed approach using real data collected by surface vehicles with SSS and demonstrate accurate flight height estimation from noisy SSS images.

Multiobject tracking (MOT) is an important task in robotics, autonomous driving, and maritime surveillance. Traditional work on MOT is model-based and aims to establish algorithms in the framework of sequential Bayesian estimation. More recent methods are fully data-driven and rely on the training of neural networks. The two approaches have demonstrated advantages in certain scenarios. In particular, in problems where plenty of labeled data for the training of neural networks is available, data-driven MOT tends to have advantages compared to traditional methods. A natural thought is whether a general and efficient framework can integrate the two approaches. This paper advances a recently introduced hybrid model-based and data-driven method called neural-enhanced belief propagation (NEBP). Compared to existing work on NEBP for MOT, it introduces a novel neural architecture that can improve data association and new object initialization, two critical aspects of MOT. The proposed tracking method is leading the nuScenes LiDAR-only tracking challenge at the time of submission of this paper.