Abstract

Mobile manipulators find applications in manufacturing, infrastructure inspection, and domestic services. While most mobile manipulators excel at pick and place, weaim to combine three distinct manipulation primitives: striking,Pushing, and pick-and-place. This integration enables versatile object manipulation and expands the robot’s operational range. Furthermore, we propose a new attachment that can be easily fitted with most industrial two-finger grippers, capable of making line contact while pushing or striking an object. The proposed framework includes a high-level action sequence planner, dedicated offline planners for each manipulation primitive (Strike, Push, and Pick & Place), and a whole-body controller. Preliminary results from simulations and hardware experiments demonstrate the potential for diverse applications in the field of mobile manipulation.

Abstract

Univariate regression trees have representation problems for non-orthogonal regression functions. Ensembles of univariate regression trees have better representational power. In some cases, weighted ensembles have shown better performance than unweighted ensembles. In this paper, we study the properties of ensembles of regression trees by using regression classification models. We propose a theoretical framework to study the representational power of infinite-sized weighted ensembles, consisting of randomised finite-sized regression trees. We show for some datasets that the weighted ensembles may have better representational power than unweighted ensembles, but the performance is highly dependent on the weighting scheme and the properties of datasets. Our model cannot be used for all the datasets. However, for some datasets, we can accurately predict the experimental results of ensembles of regression trees.

Abstract

Aim We test the hypothesis that wind dispersal is more common among emergent tree species given that being tall increases the likelihood of effective seed dispersal. Location Americas, Africa and the Asia-Pacific. Time period 1970–2020. Major taxa studied Gymnosperms and Angiosperms. Methods We used a dataset consisting of tree inventories from 2821 plots across three biogeographic regions (Americas, Africa and Asia-Pacific), including dry and wet forests, to determine the maximum height and dispersal strategy of 5314 tree species. A web search was used to determine whether species were wind-dispersed. We compared differences in tree species maximum height between biogeographic regions and examined the relationship between species maximum height and wind dispersal using logistic regression. We also tested whether emergent tree species, that is species with at least one individual taller than the 95% height percentile in one or more plots, were disproportionally wind dispersed in dry and wet forests within each biogeographic region. Results Our dataset provides maximum height values for 5314 tree species, of which more than half (2914) had no record of this trait in existing global databases. We found that, on average, tree species in the Americas have lower maximum heights compared to those in Africa and the Asia Pacific. The probability of wind dispersal increased significantly with tree species maximum height and was significantly higher among emergent than non-emergent tree species in both dry and wet forests in all three biogeographic regions. Main conclusion Wind dispersal is more prevalent in tall, emergent tree species than in non-emergent species and may thus be an important factor in the evolution of tree species maximum height. By providing the most comprehensive dataset so far of tree species maximum height and wind dispersal strategies, this study paves the way for advancing our understanding of the eco-evolutionary drivers of tree size.

Abstract

Wetlands, which serve as transitional zones between terrestrial and aquatic environments, play a crucial role in providing diverse ecosystem services for the sustenance of wildlife and mankind. Unfortunately, these delicate systems have experienced illicit human interventions in recent times, resulting in degradation and deterioration. Despite being designated as protected areas, wetlands are being exploited beyond their ecological capacity, posing a threat to their sustainability and potential existence in the near future. The present study aims to analyse the historical spatial changes in three wetlands located in Kerala, India, both before and after their declaration as Ramsar sites. Ashtamudi Lake, Sasthamkotta Lake, and Vembanad Lake were the focus areas, and the study employed multi-temporal satellite data sets from 1988 to 2023 to generate land use and land cover maps. The objective was to assess the ecological transitions, particularly the expansion of urbanization and subsequent reclamation of wetlands. By analysing the changes in aerial extent of each land cover class within the study area, the study identified spatio-temporal transitions and modifications occurring in and around the Ramsar sites. A comparison of the thematic maps from 1988 to 2023 revealed a significant increase in built-up areas and a noticeable reduction in vegetation, agriculture, and water bodies, both in terms of extent and quality. The study identified various socio-economic drivers that are responsible for these changes and emphasized the urgent need for regular monitoring and management of these wetland ecosystems. It stresses the importance of treating wetland conservation as a priority.

Abstract

The understanding of spatio-temporal variation in land use and land cover (LULC) patterns is crucial for managing catchment land use planning, as it directly influences of tropical reservoir water quality and the subsequent Nutrient Contamination (NC) of unmonitored water bodies. The current research attempts to accurately measure the influence of LULC and its associated determinants on the quantities of NC loads by using Chl-a as a proxy, within tropical reservoirs, i.e. Bhadra and Tungabhadra, located in same river catchment. This Chl-a spread calculated by the Maximum Chlorophyll Index (MCI) derived from Sentinel 2 satellite data products covering the period from July 2016 to June 2021 were done using Google Earth Engine (GEE) platform. The validation analysis confirms the robustness of the methodology with a strong correlation between MCI-calculated values and EOMAP (Earth Observation and Environmental Services Mapping) Chl-a (μg/L) data points for both reservoirs, Bhadra (R2 = 0.64) and Tungabhadra (R2 = 0.68). The findings reveal that, Tungabhadra reservoir consistently exhibits an excessive spatial distribution of Chl-a spread area (17 km2 to 335 km2 ), reflecting nutrient-rich water inflows, particularly evident during the post-monsoon period. This notable rise could be linked to harvesting the Kharif crop, resulting in elevated nutrient concentrations. In contrast Bhadra reservoir, dominated by forested areas, maintains relatively lower Chl-a spread areas (<20 km2 ), highlighting its pivotal role in maintaining water cleanliness and serves as a riparian boundary. In addition, the changes in LULC classes show a strong relationship with variation in Chl-a during the studied period, for the Bhadra Reservoir R2 = 0.51 (F- statistics = 3.983, p = 0.021), and the Tungabhadra Reservoir R2 = 0.802 (F- statistics = 7.489, p = 0.0143). This highlights how changes in land use significantly shape contamination dynamics, deepening our understanding of nutrient inputs and contamination drivers in tropical reservoirs.

Abstract

This work tackles the important task of understanding out-of-distribution behavior in two prominent types of generative models, i.e., GANs and Diffusion models. Understanding this behavior is crucial in understanding their broader utility and risks as these systems are increasingly deployed in our daily lives. Our first contribution is demonstrating that diffusion spaces outperform GANs' latent spaces in inverting high-quality OOD images. We also provide a theoretical analysis attributing this to the lack of prior holes in diffusion spaces. Our second significant contribution is to provide a theoretical hypothesis that diffusion spaces can be projected onto a bounded hypersphere, enabling image manipulation through geodesic traversal between inverted images. Our analysis shows that different geodesics share common attributes for the same manipulation, which we leverage to perform various image manipulations. We conduct thorough empirical evaluations to support and validate our claims. Finally, our third and final contribution introduces a novel approach to the few-shot sampling for out-of-distribution data by inverting a few images to sample from the cluster formed by the inverted latents. The proposed technique achieves state-of-the-art results for the few-shot generation task in terms of image quality. The effectiveness of the generated samples is further validated when used to augment data for underrepresented classes in a classification task, leading to improved performance. Our research underscores the promise of diffusion spaces in out-of-distribution imaging and offers avenues for further exploration.

Abstract

Accurately recognizing a revisited place is crucial for embodied agents to localize and navigate. This requires visual representations to be distinct, despite strong variations in camera viewpoint and scene appearance. Existing visual place recognition pipelines encode the whole image and search for matches. This poses a fundamental challenge in matching two images of the same place captured from different camera viewpoints: the similarity of what overlaps can be dominated by the dissimilarity of what does not overlap. We address this by encoding and searching for image segments instead of the whole images. We propose to use open-set image segmentation to decompose an image into `meaningful' entities (i.e., things and stuff). This enables us to create a novel image representation as a collection of multiple overlapping subgraphs connecting a segment with its neighboring segments, dubbed SuperSegment. Furthermore, to efficiently encode these SuperSegments into compact vector representations, we propose a novel factorized representation of feature aggregation. We show that retrieving these partial representations leads to significantly higher recognition recall than the typical whole image based retrieval. Our segments-based approach, dubbed SegVLAD, sets a new state-of-the-art in place recognition on a diverse selection of benchmark datasets, while being applicable to both generic and task-specialized image encoders. Finally, we demonstrate the potential of our method to revisit anything by evaluating our method on an object instance retrieval task, which bridges the two disparate areas of research: visual place recognition and object-goal navigation, through their common aim of recognizing goal objects specific to a place.

Abstract

No Results Found

Abstract

Efficiently generating grasp poses tailored to specific regions of an object is vital for various robotic manipulation tasks, especially in a dual-arm setup. This scenario presents a significant challenge due to the complex geometries involved, requiring a deep understanding of the local geometry to generate grasps efficiently on the specified constrained regions. Existing methods only explore settings involving table-top/small objects and require augmented datasets to train, limiting their performance on complex objects. We propose CGDF: Constrained Grasp Diffusion Fields, a diffusion-based grasp generative model that generalizes to objects with arbitrary geometries, as well as generates dense grasps on the target regions. CGDF uses a part-guided diffusion approach that enables it to get high sample efficiency in constrained grasping without explicitly training on massive constraint-augmented datasets. We provide qualitative and quantitative comparisons using analytical metrics and in simulation, in both unconstrained and constrained settings to show that our method can generalize to generate stable grasps on complex objects, especially useful for dual-arm manipulation settings, while existing methods struggle to do so. More results, code and an extended version of the paper can be found on the project page: https://constrained-grasp-diffusion.github.io/

Abstract

This work presents a novel approach to identifying possible early accident-prone zones in a large city-scale road network using geo-tagged collision alert data from a vehicle fleet. The alert data has been collected for a year from 200 city buses installed with the Advanced Driver Assistance System (ADAS). To the best of our knowledge, no research paper has used ADAS alerts to identify the early accident prone zones. A nonparametric technique called Kernel Density Estimation (KDE) is employed to model the distribution of alert data across stratified time intervals. A novel recall-based measure is introduced to assess the degree of support provided by our density-based approach for existing, manually determined accident-prone zones (‘blackspots’) provided by civic authorities. This shows that our KDE approach significantly outperforms existing approaches in terms of the recall-based measure. Introducing a novel linear assignment Earth Mover Distance based measure to predict previously unidentified accident-prone zones. The results and findings support the feasibility of utilizing alert data from vehicle fleets to aid civic planners in assessing accident-zone trends and deploying traffic calming measures, thereby improving overall road safety and saving lives.