Test out our new Bento Search

Automating contact-rich manipulation of viscoelastic objects with rigid robots faces challenges including dynamic parameter mismatches, unstable contact oscillations, and spatiotemporal force-deformation coupling. In our prior work, a Compliance-Aware Tactile Control and Hybrid Deformation Regulation (CATCH-FORM-3D) strategy fulfills robust and effective manipulations of 3D viscoelastic objects, which combines a contact force-driven admittance outer loop and a PDE-stabilized inner loop, achieving sub-millimeter surface deformation accuracy and ±5% force tracking. However, this strategy requires fine-tuning of object-specific parameters and task-specific calibrations, to bridge this gap, a CATCH-FORM-ACTer is proposed, by enhancing CATCH-FORM-3D with a framework of Action Chunking with Transformer (ACT). An intuitive teleoperation system performs Learning from Demonstration (LfD) to build up a long-horizon sensing, decision-making and execution sequences. Unlike conventional ACT methods focused solely on trajectory planning, our approach dynamically adjusts stiffness, damping, and diffusion parameters in real time during multi-phase manipulations, effectively imitating human-like force-deformation modulation. Experiments on single arm/bimanual robots in three tasks show better force fields patterns and thus $10\%-20\%$ higher success rates versus conventional methods, enabling precise, safe interactions for industrial, medical or household scenarios.

The use of marine protected areas (MPAs) is expanding around the world. MPAs can have a wide variety of objectives (e.g., science, conservation, food security, cultural value), and scientific guidance on how to design MPAs to achieve objectives is often based on simulation modeling. Many different models may all provide an answer to questions such as the predicted change in population biomass and fisheries catches resulting from th implementation of an MPA. When multiple levels of model complexity are all in theory capable of answering the same question, and the models cannot be confronted with data directly, the decision of what level of model complexity to use can be ad hoc. In this, paper I compare the predicted effects of MPAs on catch and biomass produced by a spatially explicit age-structured multi-species and multi-fleet (High-definition) model to the predictions generated by a two-patch surplus production (Low-definition) model, fitted to emulate the High-definition model. I found that in many cases, the predictions made by the two models were markedly different, with the Low-definition model frequently predicting substantially higher biomass benefits from MPAs than the High-definition model, and in some cases incorrectly estimating the direction (positive or negative) of the MPA effects. However, I also found that the Low-definition model has strategic value for broad classification and ranking exercises. My results show that care should be taken in selecting and interpreting the results of MPA simulation models and that research is needed to understand what models are best suited to what policy recommendations when multiple viable options exist.

Spillover is a term commonly applied to the dispersal of fish and/or larvae from inside a closed area to areas open to fishing. The presence of spillover is often quantified by measuring gradients in attributes such as abundance or catch rates near the boundaries of closed areas or by measuring higher abundance inside closed areas compared to outside. It is commonly assumed that such gradients or ratios indicate that the closed area has benefitted the fishery and the total abundance of fish. We explore this assumption using a spatially explicit model of closed areas with different intensities of fishing and fish movement, and we find that such gradients will be expected any time there is higher abundance inside the closed area. However, such gradients do not necessarily indicate a benefit to the fishery either in terms of total catch or catch rate, and unless pre-closure fishing was intense, total abundance is not expected to rise significantly. We examine case studies that argue that spillover exists and leads to fishery benefits. We then evaluate the evidence for net benefits in these case studies and find those with evidence of net benefits all come from places where fishing pressure was intense. While most analysis come from quite small coastal closed areas, two studies of very large open-ocean closed areas are discussed, and we find that both suggest little overall impact on the tuna populations that support the main commercial fisheries affected by the closures in question.

Fish welfare is a crucial issue that needs to be addressed in fisheries. Thus, the scope of the fair-fish database - an online open-access platform - was expanded from aquaculture (farm branch) to fisheries (catch branch). It provides farm and catch welfare profiles (WelfareChecks) of aquatic species based on literature reviews. In the catch branch, each WelfareCheck encompasses a species in relation to a specific fishing method used to catch it, assessing 10 criteria covering welfare hazards throughout the steps of the catching process: prospection, setting, catching, emersion, release from gear, bycatch avoidance, sorting, discarding, storing, and stunning/slaughter. In each criterion, we assess the likelihood and potential of experiencing good welfare under minimal and high-standard fisheries conditions, respectively, besides the certainty level about these. A final WelfareScore is provided for each profile, which serves as a benchmark for assessing and improving fish welfare. Since its publication in 2023, we have published five WelfareChecks. The goal is to increase the number of profiles for several fished species and catching methods over time. In conclusion, the catch branch of the fair-fish database serves as an open-access source providing an overview of the welfare of a fished species given a certain catching method. It is a reliable tool that raises public awareness of fish welfare, provides scientists with insight into knowledge gaps, and offers practitioners with suggestions about how to avoid welfare risks.

Capture in nearshore fisheries is the leading threat to coastal elasmobranchs, of which more than 75 % are threatened with extinction globally. Limited knowledge of these highly dynamic fisheries impedes the design and implementation of stakeholder-inclusive policies for conservation. To address this, we developed an interdisciplinary approach, combining landing data with fishing geo-locations, Very High resolution (VHR) satellite imagery and fisher interviews to model elasmobranch catch dynamics and map areas of high catch potential. We compared how elasmobranch catch rates varied by species ecology, habitat, and fisheries characteristics in Visakhapatnam and Malvan, two regions on the east and west coasts of India, respectively. We sampled 2209 fishing trips across three oceanographic seasons from landing sites at both locations in 2022-23. We recorded 5578 elasmobranchs from >20 species of which at least 13 were categorised as ‘Threatened’. Gillnets, hook and line and trawl nets were the most common gears, but their use and catch rates varied considerably. Elasmobranchs had a higher catch risk on the eastern site (where they may be specifically targeted) and were generally larger. Catch rates were higher in shallow regions on the west coast and in the summer at both sites. Importantly, we demonstrate that drivers of elasmobranch catch were site and fishery specific, underscoring the need for more local-scale research for planning conservation actions. Our framework provides a robust method to study the highly dynamic and diverse nature of nearshore fisheries, which can inform conservation actions and, at the same, time, enable a bottom-up approach to conserving elasmobranchs.

Accurate disease detection in fruit leaves is crucial for preserving crop health and enhancing agricultural productivity. Fruits hold significant economic value and nutritional importance, but their vulnerability to diseases severely impacts both yield and quality. This research proposes a novel Multi-level Attention DenseNet-based Deep Convolutional Neural Network (MADDCNN) model to improve fruit disease detection. The model addresses challenges such as varying lighting conditions, complex backgrounds, and overlapping symptoms, which hinder detection accuracy in existing systems. By incorporating the Catch Fish Optimization (CFO) technique, MADDCNN enhances multi-fruit classification accuracy and optimizes parameter tuning for better performance. To further enhance the detection process, the model employs the Hybrid Runge Kutta DeepLabV3+ (HRKD) method for segmenting disease-affected areas, enabling more precise identification and isolation of infected regions. This segmentation step significantly boosts classification accuracy and reliability in agricultural environments. The novelty of this research lies in the integration of MADDCNN with CFO and HRKD, enabling severity classification, adaptive optimization, and precise segmentation under varying agricultural conditions. The MADDCNN model outperforms other methods in key metrics, achieving 98.34% accuracy, 97.85% precision, 97.91% recall, and 98.12% F1 score. Furthermore, it demonstrates computational efficiency, processing each image in just 1.5 s. Overall, the MADDCNN approach offers a sustainable, efficient solution for detecting fruit diseases, addressing the challenges of varying conditions and complex symptoms, and contributing to enhanced agricultural practices.