Publication type: Conferences
A crucial task in marine research is that of monitoring fish in order to quantify different species and analyze their behavior. Computer Vision clearly helps in solving this problem, especially when combined with Deep Learning. This paper focuses on the use of Deep Learning and Computer Vision to perform fish detection and tracking. More specifically, in this study we survey different state-of-the-art object detectors, we adapt and train them to perform fish detection, and we evaluate and compare them by means of a large underwater video sequence involving fish in their natural habitat. The experimental results, provided in terms of detection quality and time consumption, will help researchers to choose the appropriate object detection model depending on the pursued goals.
Real-time image exchange among different marine autonomous robots can be a requirement in some applications, and it becomes a real challenge particularly in underwater systems with restricted bandwidth communications. Some examples where visual data transmission is needed include: a) detecting inter-session visual loop closings in multi-robot SLAM approaches, b) remote and fast awareness of underwater biologic or archaeological hot-spots from mobile submersed robots to any surfaced station and from this to the world or c) cooperative intervention using a robotic team demanding different points of view of the elements of interests to be manipulated, taken from different autonomous robots equipped with cameras. Since optical links imply high directionality and limited ranges, acoustic modems turn out to be the most used transmission system for their versatility and omnidirec- tionality, despite their very limited bandwidth. Dealing with bandwidth limitations in underwater omni-directional acoustic links has been an important objective in applications that require loss-less and efficient data exchange, for instance, in cooperative sampling or localization, but scarcely approached to transmit entire images, rather than their corresponding visual features of global signatures, in multi-agent intervention systems.
This paper proposes a method to progressively encode underwater images. The image information is split into several, very small, parts called chunks that can be easily transmitted using unreliable, low bandwidth, underwater communication channels. The receiver can reconstruct the original image with increasing quality after each chunk is received up to a perfect, lossless, reconstruction.
This paper presents an algorithm to reject false loops ready to be used as a front-end for GraphSLAM. The proposal operates in two steps. The first one checks each loop independently to reduce the computational cost of the second one, which jointly checks the consistency of several loops. The experiments show the ability of this algorithm to detect wrong loops, favourably comparing to a previously existing approach in terms of final graph quality and time consumption.
In the last years visual odometers have been improved considerably, but there is
still a certain lack of robustness and reliability when used to navigate Autonomous Underwater
Vehicles (AUV) in complex underwater scenarios, such as habitats colonized with Posidonia
oceanica or Zostera noltii. The work presented in this paper goes one step beyond the current
solutions, improving a state of the art approach to be applied in underwater applications. Tests
conducted with real marine visual data grabbed in waters of the Balearic Islands from a bottom-looking camera installed on an AUV, show the progress in the vehicle displacement estimation and its viability to be used online.
The integration of Time-Sensitive Networks (TSN) with 5G cellular networks requires a defined architecture for network configuration and management. Although 3GPP specifications provide necessary means for the TSN-5G integration, the operation of such converged TSN-5G network remains an open challenge for the research community. To address this challenge, this paper presents the ongoing work in developing a centralized architectural model to configure the TSN-5G network, and forward traffic from TSN to 5G and vice-versa. The proposed architectural model uses knowledge of the traffic characteristics to carry out a more accurate mapping of quality of service attributes between TSN and 5G.
DOI: https://doi.org/10.1109/ETFA52439.2022.9921731
In this paper, we present our ongoing work on proposing solutions to integrate legacy end-stations into Time-Sensitive Network (TSN) communication systems where the legacy end-stations are synchronized via their legacy clock synchronization protocol. To this end, we experimentally identify the effects of lacking synchronization or partial synchronization in TSN networks. In the experiments we show the effects of clock synchronization in different scenarios on jitter and clock drifts. Based on the experiments, we propose preliminary solutions to overcome the identified effects.
DOI: https://doi.org/10.1109/ETFA52439.2022.9921709
Novel industrial applications are leading to important changes in industrial systems. One of the most important changes is the need for systems that are capable to adapt to changes in the environment or the system itself. Because of their nature many of these applications are distributed, and their network infrastructure is key to guarantee the correct operation of the overall system. Furthermore, in order for a distributed system to be able to adapt, its network must be flexible enough to support changes in the traffic during runtime. The Time-Sensitive Networking (TSN) Task Group has proposed a series of standards that aim at providing deterministic real-time communications over Ethernet. TSN also provides centralised online configuration and control architectures which enable the online configuration of the network. A key part in TSN’s centralised architectures is the Centralised Network Configuration element (CNC). In this work we present a first implementation of a CNC capable of scheduling time-triggered traffic and deploying such configuration in the network using the Network Configuration (NETCONF) protocol. We also assess the correctness of our implementation using an industrial use case provided by Volvo Construction Equipment.
DOI: https://doi.org/10.1109/ETFA52439.2022.9921518