Publications

Using industrial by-products as substitutes for Ordinary Portland Cement (OPC) is a promising strategy to reduce its environmental impact. However, heavy metals like Pb strongly interfere with initial kinetics. The dynamic physicochemical environment makes it challenging to identify the key factors. Here, we employed in-situ XRD as a time-dependent method, alongside conventional characterization techniques and geochemical modeling, to investigate the Pb-induced retardation in CEMI 42.5R and 52.5R. The results show that Pb-hydroxides and Pb-O-Si clusters are expected to be the primary mechanisms for this inhibition. Among clinker phases, C3A dissolution is less affected and serves as the primary source of alkalinity in early hydration, promoting hydration products precipitation and gypsum dissolution. Geochemical modeling suggests that Pb species concentration in the solution regulates the precipitation of hydration products, especially portlandite. The comparison of hydration kinetics of 2 types of OPC highlights optimizing particle size as a solution to mitigate retardation impact.

In recent years, there has been a significant interest in developing machine learning algorithms on embedded systems. This is particularly relevant for bare metal devices in Internet of Things, Robotics, and Industrial applications that face limited memory, processing power, and storage, and which require extreme robustness. To address these constraints, we present MicroFlow, an open-source TinyML framework for the deployment of Neural Networks (NNs) on embedded systems using the Rust programming language. The compiler-based inference engine of MicroFlow, coupled with Rust’s memory safety, makes it suitable for TinyML applications in critical environments. The proposed framework enables the successful deployment of NNs on highly resource-constrained devices, including bare-metal 8-bit microcontrollers with only 2 kB of RAM. Furthermore, MicroFlow is able to use less Flash and RAM memory than other state-of-the-art solutions for deploying NN reference models (i.e. wake-word and person detection), achieving equally accurate but faster inference compared to existing engines on medium-size NNs, and similar performance on bigger ones. The experimental results prove the efficiency and suitability of MicroFlow for the deployment of TinyML models in critical environments where resources are particularly limited.

We propose a flexible hierarchical navigation stack for a mobile robot in complex dynamic environments. Addressing the growing need for reliable navigation in real-world scenarios, where dynamic agents and environmental uncertainties pose significant challenges, our solution decomposes this complexity into task planning, navigation, control, and safe velocity components. In contrast to the prior art, our system at every level incorporates diverse contextual information about the environment, anticipates navigation risks and proactively avoids collisions with dynamic agents.

Ordinary Portland cement (OPC) is a ubiquitous construction material and has long been the most prevalent of all man-made concepts. However, the massive demand for OPC is responsible for approximately 7–8% of all anthropogenic CO2 emissions. Substituting OPC with industrial by-products presents a promising avenue for reducing clinker usage and aiding industry decarbonization. However, concerns arise regarding the presence of trace metals, particularly Pb, which can impede early hydration and degrade material properties. Understanding the kinetics of clinker phase dissolution in the presence of Pb is crucial for mitigating these issues. Conventional characterization methods may alter samples and fail to adequately capture underlying reaction mechanisms. To address this challenge, our study employs in-situ X-ray diffraction (XRD) to accurately assess Pb-OPC hydration kinetics in real time. Furthermore, we develop a geochemical model to quantify hydration reactions. This model supplements experimental findings, providing valuable insights into the proposed mechanisms. Overall, our work enhances the understanding of Pb-OPC interactions in cementitious materials, ultimately contributing to more efficient industrial by-product management and sustainable construction practices.

Dense annotation of real 3D LiDAR point clouds for mobile robot applications remains challenging. Unsupervised Domain Adaptation (UDA) enables the segmentation of unlabeled real-world point clouds by leveraging labeled synthetic data. However, existing self-training-based UDA methods rely on fixed thresholds for pseudo-label selection, limiting adaptation performance. In this work, we address this limitation. We propose a novel UDA framework for 3D LiDAR semantic segmentation, centered on a confidence-guided pseudo-label sampling strategy (ConSamp). Specifically, ConSamp adopts a probabilistic sampling strategy in which pseudo-labels with higher confidence are more likely to be retained. Meanwhile, the sampling function itself evolves adaptively throughout training to respond to changes in confidence distribution. Experiments show that our model achieves strong performance on synthetic-to-real 3D LiDAR semantic segmentation tasks. In particular, results better than state-of-the-art methods have been achieved on two public 3D point cloud datasets: SemanticKITTI [1] and SemanticPOSS [2].

Automatic fruit growth monitoring plays a vital role in advancing precision agriculture. Tracking the evolution of fruits over time is essential to monitor their development and optimize production. The ability to recognize fruits over periods of time, even with drastic scene changes, is a required capability of agricultural robots. This letter presents a system that allows long-term fruit tracking in 3D data. It generates instance-segmented 3D representations of plants at various growth stages over time, utilizing only consumer-grade RGB-D cameras installed on a mobile robot. Our approach first performs instance segmentation on each image in a sequence. Then, by exploiting geometric information and depth maps, we track the same instances throughout the sequence. We produce a 3D point cloud containing instances, exploiting odometry information and 3D semantic mapping. Once our robot performs a new recording at a different plant growth stage, it associates each fruit with the previously built 3D cloud and update the model. We validate the system in a real-world glasshouse environment in Bonn, Germany. Experimental results demonstrate that our system outperforms existing baselines even though it relies only on annotated images and operates at frame-rate, allowing the deployment on a real robot.

One of the biggest challenges in autonomous underwater navigation is the capability of the autonomous underwater vehicle (AUV) to localize itself, since common positioning systems (e.g., GPS or USBL), when available, can be unstable and very noisy. In this paper, we address the problem of place recognition in underwater synthetic and real environments, which is a key component in autonomous localization for robotics and navigation systems. In underwater scenarios, cameras are often subject to water turbidity and low-light conditions, making their use unreliable. Sonar data on the other hand is not affected by these limitations, but its interpretation is more challenging. In this paper we introduce a global descriptor for multibeam sonar images, to be compared with a database of sonar image descriptors acquired at known locations in sparsely structured environments. To enforce the similarity between descriptors computed from nearby poses, we introduce a novel loss that correlates the oriented-Intersection over Union (o-IoU) between pairs of sonar scans with the corresponding distances between their descriptors. A proxy image reconstruction loss has also been integrated for self-supervised adaptation to real data. Preliminary experimental results show that our method is able to localize an AUV in real-time in both synthetic and real environments by training it for localization using only synthetic sonar images.