Cycle 37
A recent work studied the biaxial static yielding behavior of specimens made of EN-GJS-600–3 ductile cast iron (DCI). The yield domain deviates from the Von Mises criterion and takes a cusp-like shape. This evidence is investigated by finite element (FE) simulations. The numerical results are in accordance with the experimental ones.
See the poster

Cycle 35
The laser technology, due to its advantages like high precision and speed, is widely used for material processing such as cutting, welding, additive manufacturing, etc. In this study, thermal and thermomechanical analysis of mild structural steel under laser cutting process is investigated.
See the poster

Cycle 36
The ability to diagnose faults is an important property of safety-critical systems. We define the fault diagnosability problem using alarm conditions that describe how quickly a fault is detected. We show that this problem amounts to finding so-called critical pairs and ribbon-shaped critical pairs.
See the poster

Cycle 37
We propose a Personal Healthcare Agent to help people with Autism Spectrum Disorder in dealing with social and emotional daily challenges by retrieving and proposing solutions as natural language recipes. The Continual Learning framework allows the PHA to acquire new knowledge from human experts.
See the poster

Cycle 34
Human Pose Estimation aims at retrieving the 3D position of human joints from images or videos. HPE methods suffer a lack of viewpoint equivariance, failing to deal with viewpoints unseen at training time. To tackle this issue we propose DECA, a novel capsule autoencoder with fast capsule routing.
See the poster

Cycle 34
We motivate why the science of learning to reject model predictions is central to ML and show how current accuracy metrics are misaligned with the value and cost induced from the AI workflows in typical enterprise scenarios.
See the poster

Cycle 36
Reduced data size leads to faster computation, transfer, and efficient storage. This study evaluates the performance of deep learning algorithm over lung ultrasound data with varying pixel and grey-level resolution. Prognostic agreement of 73.5% can be achieved with 32 times reduced data size.
See the poster

Cycle 35
This work considers an ultra-dense edge network with MEC facilities for optimally managing task offloading in the context of heterogeneous computing and communication services required by real-time robotic applications. The model is formalized as MILP to obtain an optimal schedule and maximum QoS.
See the poster

Cycle 36 – 35
To foster reusable data creation, we propose iTelos, a data collection-integration methodology. iTelos leads the data collection with the iLog app and builds reusable Knowledge Graph adapting existing data to contextual requirements. iTelos has been used within university courses and EU projects.
See the poster

Cycle 34
The performance of image-based 3D reconstruction algorithms largely depends on the image quality and the camera network. From an end-user perspective, it is hard to monitor these aspects during the acquisition of the images. Leveraging recent visual SLAM developments, we present a system for enabling assisted and more controlled image acquisitions.
See the poster

Cycle 36
Even though general-purpose multicore processors provide a high degree of parallelism and programming flexibility intermittent computing has not exploited them yet. We present a runtime that supports parallel intermittent computing and adjusts the underlying multicore architecture to ambient power.
See the poster

Cycle 35
Weighted First-Order Model Counting (WFOMC) serves as a workhorse for many Probabilistic Graphical Models (PGMs). In this work, we provide a closed-form formula for polynomial-time WFOMC for a large class of First-Order Logic theories, allowing efficient exact inference in large families of PGMs.
See the poster

Cycle 34
The work addresses the problem of multi-robot navigation in human shared spaces. We propose a hierarchical framework that combines global path planning, local path planning, and reactive strategies, ensuring safe and socially-aware navigation even in challenging scenarios with multiple humans.
See the poster