SMaRC Academy Seminars, May 7th 2021

Tracking and Forecasting of Algal Blooms Using Satellite Ocean Data and USVs

Abstract: Harmful algal blooms occur frequently and cause human illness, large-scale mortality of fish, shellfish, mammals, and birds, and deteriorates water quality. To control these detrimental effects, accurate information about the location and movement patterns of algal blooms are valuable. An indicator of algal bloom presence is chlorophyll concentration.

We consider the problem of tracking algal blooms front using a USV and satellite data. The USV has a sensor that measures the concentration of chlorophyll. We propose a protocol that allows the USV to find and track algal bloom fronts by least-squares estimation from past chlorophyll measurements. The satellite provides past data from the Baltic sea on chlorophyll, salinity, temperature, and other relevant parameters. We use neural networks to forecast algal bloom spread, given by chlorophyll concentration. Given the algal bloom forecast from the satellite and the local chlorophyll measurements from the USV, we use Gaussian regression to reduce the ocean sampling error.

Current implementation efforts include Matlab simulations, Neptus/Dune numerical experiments, and drone experiments at SML. The next step is sea/ocean implementation using the Neptus/Dune platform.

Autonomous systems and the implementation of innovative technology

Abstract: Autonomous systems are researched, developed, and debated in many different domains, states and entities, the Armed Forces included. The aim for this PhD work is to sharpen the ability within the Swedish Armed Forces in general, the Navy in particular, to understand the technical development of autonomous systems in relation to other research areas.

Three studies will be presented in short. The first one handle issues concerning the concept development of innovative technologies into organisations like armed forces, the second one focus on the ethical and legal aspects of using autonomous underwater systems, especially in a military context. The third study is a case study where different units within the Swedish Armed Forces will be questioned on their participation in capability development when introducing new technology. The presentation will end with two examples from the studies: how the capability perspective could be used in concept development and how rules and ethical aspects can be related to the technological systems.

AUV Localization with Side-Scan Sonars

Abstract: In underwater environments, the perception and navigation systems are heavily dependent on the acoustic wave based sonar technology. Sidescan sonar (SSS) provides high-resolution, photo-realistic images of the seafloor at a relatively cheap price. These images could be considered potential candidates for place recognition and navigation of autonomous underwater vehicles (AUVs).
In this talk, I will describe the characteristics of sidescan sonar, the benefits as well as challenges with using sidescan sonar data for AUV localization. Specifically, I will present a deep learning based approach to SSS image correspondence matching, where SSS images from a seafloor area with bottom trawling marks were used.

Lastly, I will briefly touch upon how sidescan data will be used in the upcoming demo at Kristineberg.

Noncoherent Acoustic Underwater Communication

Abstract: The underwater domain poses many difficulties for any communicating platform; water attenuates signals that are transmitted using radio, light, and sound. However, the attenuation of sound is highly dependent on frequency, and can be used to communicate over kilometers using a limited bandwidth.  The acoustic channel often has a relatively long and time-varying impulse response, and tracking the phase shift due to the channel is a difficult and energy-costly problem to solve for a coherent method. A noncoherent method does not require phase tracking, and is inherently more robust. This talk will showcase the varietly of underwater channels, and how a noncoherent method deals with them.

SMaRC Academy Seminars, May 17th 2021

Multi-AUV Mission planning and execution

Abstract:In this seminar, I will present my work aiming at using multiple AUVs in order to increase the effectiveness of the mission and even enable missions that are otherwise impossible with a single vehicle. First I will present my work on creating effective 3D cages underwater that are constructed using the sensing capabilities of many AUVs such that an arbitrary evader can not escape. I will show how taking into account the geography can enable even larger volumes of water to be caged effectively. I will then talk about how we can control many AUVs in close proximity in a safe manner while keeping mission completion guarantees using Control Barrier Functions and Behavior Trees. Then I will move into coverage path planning under localization uncertainties and how we can plan for non-gaussian localization errors and still guarantee area coverage. Finally I will talk about my current work on cooperative coverage with multiple AUVs over very large areas and how we can utilize the system of AUVs to improve the localization accuracy of individual AUVs using Pose Graphs.

Robotic Perception for Autonomous Underwater Navigation and Mapping

Abstract: Robotic perception is the first step in the pipeline when developing an autonomous system. Feedback from different types of sensors is paramount for autonomous navigation, In contrast with surface, ground, and aerial mobile robots, which can use, e.g., optical and laser sensors, underwater robots rely mostly on acoustic devices to understand and navigate their surroundings. However, the integration of the sensors and interpretation of the data on AUVs and ASVs does not go without hassle. This presentation will show some of the challenges and results on the work to integrate multibeam sonars for autonomous underwater navigation and mapping.

High-Resolution Bathymetric Reconstruction From Sidescan Sonar With Deep Neural Network

Abstract: This talk will present our work of proposing a novel data-driven approach for high-resolution bathymetric reconstruction from sidescan. Sidescan sonar (SSS) intensities as a function of range do contain some information about the slope of the seabed. However, that information must be inferred. Additionally, the navigation system provides the estimated trajectory, and normally the altitude along this trajectory is also available. From these, a very coarse seabed bathymetry can be obtained as an input. This coarse bathymetry is then combined with the indirect but high-resolution seabed slope information from the sidescan to estimate the full bathymetry. This sparse depth could be acquired by single-beam echo sounder, Doppler Velocity Log (DVL), other bottom tracking sensors or bottom tracking algorithm from sidescan itself. A fully convolutional network is used to estimate the depth contour and its aleatoric uncertainty from the sidescan images and sparse depth in an end-to-end fashion. The estimated depth is then used together with the range to calculate the point’s 3D location on the seafloor. I will present that a high-quality bathymetric map can be reconstructed after fusing the depth predictions and the corresponding confidence measures from the neural networks. I will show the improvement of the bathymetric map gained by using sparse depths with sidescan over estimates with sidescan alone. I will also show the benefit of confidence weighting when fusing multiple bathymetric estimates into a single map. Finally, I would discuss the future work of improving the map in fine details by introducing another way to represent the map.

Task execution with behaviour trees

Abstract: Behaviour trees (BT) are mathematical models of task execution plans. In comparison to finite-state machines, they have the advantages of modularity — allowing components to be easily replaced and reused — and a deliberative structure — enabling goal-orientated behaviors. Due to these advantages, we have made heavy use of BTs in the SMaRC project to accomplish high-level goals with our AUVs in scenarios such as target searching and environmental monitoring. To achieve high-level goals, a BT delegates the use of low-level controllers, which are often inherently continuous in time. Despite this, virtually all formulations of BTs have been in discrete time. In this talk, I will present some of our recent results on formulating BTs in continuous time. Additionally, I will present some previous related results.

AUV LoLo