Lorenzo Natale

Zhichao Li, Chiara Bartolozzi, Lorenzo Natale et al.

Object pose tracking is a fundamental and essential task for robotics to perform tasks in the home and industrial settings. The most commonly used sensors to do so are RGB-D cameras, which can hit limitations in highly dynamic environments due to motion blur and frame-rate constraints. Event cameras have remarkable features such as high temporal resolution and low latency, which make them a potentially ideal vision sensors for object pose tracking at high speed. Even so, there are still only few works on 6D pose tracking with event cameras. In this work, we take advantage of the high temporal resolution and propose a method that uses both a propagation step fused with a pose correction strategy. Specifically, we use 6D object velocity obtained from event-based optical flow for pose propagation, after which, a template-based local pose correction module is utilized for pose correction. Our learning-free method has comparable performance to the state-of-the-art algorithms, and in some cases out performs them for fast-moving objects. The results indicate the potential for using event cameras in highly-dynamic scenarios where the use of deep network approaches are limited by low update rates.

Tommaso Galliena, Stefano Rosa, Tommaso Apicella et al.

Vision-Language Models (VLMs) often yield inconsistent descriptions of the same object across viewpoints, hindering the ability of embodied agents to construct consistent semantic representations over time. Previous methods resolved inconsistencies using offline multi-view aggregation or multi-stage pipelines that decouple exploration, data association, and caption learning, with limited capacity to reason over previously observed objects. In this paper, we introduce a unified, memory-augmented Vision-Language agent that simultaneously handles data association, object captioning, and exploration policy within a single autoregressive framework. The model processes the current RGB observation, a top-down explored map, and an object-level episodic memory serialized into object-level tokens, ensuring persistent object identity and semantic consistency across extended sequences. To train the model in a self-supervised manner, we collect a dataset in photorealistic 3D environments using a disagreement-based policy and a pseudo-captioning model that enforces consistency across multi-view caption histories. Extensive evaluation on a manually annotated object-level test set, demonstrate improvements of up to +11.86% in standard captioning scores and +7.39% in caption self-similarity over baseline models, while enabling scalable performance through a compact scene representation. Code, model weights, and data are available at https://github.com/hsp-iit/epos-vlm

Federico Ceola, Elisa Maiettini, Giulia Pasquale et al.

The visual system of a robot has different requirements depending on the application: it may require high accuracy or reliability, be constrained by limited resources or need fast adaptation to dynamically changing environments. In this work, we focus on the instance segmentation task and provide a comprehensive study of different techniques that allow adapting an object segmentation model in presence of novel objects or different domains. We propose a pipeline for fast instance segmentation learning designed for robotic applications where data come in stream. It is based on an hybrid method leveraging on a pre-trained CNN for feature extraction and fast-to-train Kernel-based classifiers. We also propose a training protocol that allows to shorten the training time by performing feature extraction during the data acquisition. We benchmark the proposed pipeline on two robotics datasets and we deploy it on a real robot, i.e. the iCub humanoid. To this aim, we adapt our method to an incremental setting in which novel objects are learned on-line by the robot. The code to reproduce the experiments is publicly available on GitHub.

Andrea Maracani, Raffaello Camoriano, Elisa Maiettini et al.

Fine-tuning and Domain Adaptation emerged as effective strategies for efficiently transferring deep learning models to new target tasks. However, target domain labels are not accessible in many real-world scenarios. This led to the development of Unsupervised Domain Adaptation (UDA) methods, which only employ unlabeled target samples. Furthermore, efficiency and privacy requirements may also prevent the use of source domain data during the adaptation stage. This challenging setting, known as Source-Free Unsupervised Domain Adaptation (SF-UDA), is gaining interest among researchers and practitioners due to its potential for real-world applications. In this paper, we provide the first in-depth analysis of the main design choices in SF-UDA through a large-scale empirical study across 500 models and 74 domain pairs. We pinpoint the normalization approach, pre-training strategy, and backbone architecture as the most critical factors. Based on our quantitative findings, we propose recipes to best tackle SF-UDA scenarios. Moreover, we show that SF-UDA is competitive also beyond standard benchmarks and backbone architectures, performing on par with UDA at a fraction of the data and computational cost. In the interest of reproducibility, we include the full experimental results and code as supplementary material.

Gianluca Scarpellini, Stefano Rosa, Pietro Morerio et al.

When an object detector is deployed in a novel setting it often experiences a drop in performance. This paper studies how an embodied agent can automatically fine-tune a pre-existing object detector while exploring and acquiring images in a new environment without relying on human intervention, i.e., a fully self-supervised approach. In our setting, an agent initially learns to explore the environment using a pre-trained off-the-shelf detector to locate objects and associate pseudo-labels. By assuming that pseudo-labels for the same object must be consistent across different views, we learn the exploration policy Look Around to mine hard samples, and we devise a novel mechanism called Disagreement Reconciliation for producing refined pseudo-labels from the consensus among observations. We implement a unified benchmark of the current state-of-the-art and compare our approach with pre-existing exploration policies and perception mechanisms. Our method is shown to outperform existing approaches, improving the object detector by 6.2% in a simulated scenario, a 3.59% advancement over other state-of-the-art methods, and by 9.97% in the real robotic test without relying on ground-truth. Code for the proposed approach and baselines are available at https://iit-pavis.github.io/Look_Around_And_Learn/.

Haodong Zheng, Gabriele M. Caddeo, Andrei C. Jalba et al.

We present an active tactile exploration framework for joint object recognition and 6D pose estimation. The proposed method integrates wrist force/torque sensing, GelSight tactile sensing, and free-space constraints within a Bayesian inference framework that maintains a belief over object class and pose during active tactile exploration. By combining contact and non-contact evidence, the framework reduces ambiguity and improves robustness in the joint class-pose estimation problem. To enable efficient inference in the large hypothesis space, we employ a customized particle filter that progressively samples particles based on new observations. The inferred belief is further used to guide active exploration by selecting informative next touches under reachability constraints. For effective data collection, a motion planning and control framework is developed to plan and execute feasible paths for tactile exploration, handle unexpected contacts and GelSight-surface alignment with tactile servoing. We evaluate the framework in simulation and on a Franka Panda robot using 11 YCB objects. Results show that incorporating tactile and free-space information substantially improves recognition and pose estimation accuracy and stability, while reducing the number of action cycles compared with force/torque-only baselines. Code, dataset, and supplementary material will be made available online.

Gianluca Scarpellini, Stefano Rosa, Pietro Morerio et al.

Object detectors often experience a drop in performance when new environmental conditions are insufficiently represented in the training data. This paper studies how to automatically fine-tune a pre-existing object detector while exploring and acquiring images in a new environment without relying on human intervention, i.e., in a self-supervised fashion. In our setting, an agent initially explores the environment using a pre-trained off-the-shelf detector to locate objects and associate pseudo-labels. By assuming that pseudo-labels for the same object must be consistent across different views, we devise a novel mechanism for producing refined predictions from the consensus among observations. Our approach improves the off-the-shelf object detector by 2.66% in terms of mAP and outperforms the current state of the art without relying on ground-truth annotations.

Carlotta Sartore, Mohamed Elobaid, Lorenzo Rapetti et al.

Collaboration is central to human behavior, enabling tasks beyond individual capability. This ability arises from coordinating actions through internal representations of others, a concept known as shared intelligence. Additionally, humans are characterized by physical bodies and cognitive abilities that are optimized in response to their environment, a phenomenon referred to as embodied cognition. Designing humanoid robots that collaborate safely and effectively with people requires unifying these principles. Here we propose an architecture that integrates shared intelligence and embodied cognition to enable robots to physically collaborate with humans, where robot hardware and control are optimized for human metrics, using representations of the human body and motion intelligence. The ultimate goal is to achieve a form of shared embodied intelligence. Specifically, our architecture optimizes robot hardware and physical intelligence parameters with respect to human ergonomic metrics. This is accomplished by modeling human-robot interaction as a function of hardware configurations and embedding human models into the robot's physical intelligence. As a concrete implementation, we present the humanoid robot ergoCub, whose morphology and control have been optimized for collaborative tasks with humans. Our approach provides a framework for designing humanoid robots that prioritize human ergonomics at both the hardware and physical intelligence levels, with applications in industrial and assistive robotics.

Federico Vasile, Elisa Maiettini, Giulia Pasquale et al.

We consider the task of object grasping with a prosthetic hand capable of multiple grasp types. In this setting, communicating the intended grasp type often requires a high user cognitive load which can be reduced adopting shared autonomy frameworks. Among these, so-called eye-in-hand systems automatically control the hand pre-shaping before the grasp, based on visual input coming from a camera on the wrist. In this paper, we present an eye-in-hand learning-based approach for hand pre-shape classification from RGB sequences. Differently from previous work, we design the system to support the possibility to grasp each considered object part with a different grasp type. In order to overcome the lack of data of this kind and reduce the need for tedious data collection sessions for training the system, we devise a pipeline for rendering synthetic visual sequences of hand trajectories. We develop a sensorized setup to acquire real human grasping sequences for benchmarking and show that, compared on practical use cases, models trained with our synthetic dataset achieve better generalization performance than models trained on real data. We finally integrate our model on the Hannes prosthetic hand and show its practical effectiveness. We make publicly available the code and dataset to reproduce the presented results.

Andrea Rosasco, Stefano Berti, Fabrizio Bottarel et al.

Many robotic tasks involving some form of 3D visual perception greatly benefit from a complete knowledge of the working environment. However, robots often have to tackle unstructured environments and their onboard visual sensors can only provide incomplete information due to limited workspaces, clutter or object self-occlusion. In recent years, deep learning architectures for shape completion have begun taking traction as effective means of inferring a complete 3D object representation from partial visual data. Nevertheless, most of the existing state-of-the-art approaches provide a fixed output resolution in the form of voxel grids, strictly related to the size of the neural network output stage. While this is enough for some tasks, e.g. obstacle avoidance in navigation, grasping and manipulation require finer resolutions and simply scaling up the neural network outputs is computationally expensive. In this paper, we address this limitation by proposing an object shape completion method based on an implicit 3D representation providing a confidence value for each reconstructed point. As a second contribution, we propose a gradient-based method for efficiently sampling such implicit function at an arbitrary resolution, tunable at inference time. We experimentally validate our approach by comparing reconstructed shapes with ground truths, and by deploying our shape completion algorithm in a robotic grasping pipeline. In both cases, we compare results with a state-of-the-art shape completion approach.

Gabriele M. Caddeo, Nicola A. Piga, Fabrizio Bottarel et al.

In this paper, we address the problem of estimating the in-hand 6D pose of an object in contact with multiple vision-based tactile sensors. We reason on the possible spatial configurations of the sensors along the object surface. Specifically, we filter contact hypotheses using geometric reasoning and a Convolutional Neural Network (CNN), trained on simulated object-agnostic images, to promote those that better comply with the actual tactile images from the sensors. We use the selected sensors configurations to optimize over the space of 6D poses using a Gradient Descent-based approach. We finally rank the obtained poses by penalizing those that are in collision with the sensors. We carry out experiments in simulation using the DIGIT vision-based sensor with several objects, from the standard YCB model set. The results demonstrate that our approach estimates object poses that are compatible with actual object-sensor contacts in $87.5\%$ of cases while reaching an average positional error in the order of $2$ centimeters. Our analysis also includes qualitative results of experiments with a real DIGIT sensor.

Gabriele M. Caddeo, Andrea Maracani, Paolo D. Alfano et al.

In this paper, we address the Sim2Real gap in the field of vision-based tactile sensors for classifying object surfaces. We train a Diffusion Model to bridge this gap using a relatively small dataset of real-world images randomly collected from unlabeled everyday objects via the DIGIT sensor. Subsequently, we employ a simulator to generate images by uniformly sampling the surface of objects from the YCB Model Set. These simulated images are then translated into the real domain using the Diffusion Model and automatically labeled to train a classifier. During this training, we further align features of the two domains using an adversarial procedure. Our evaluation is conducted on a dataset of tactile images obtained from a set of ten 3D printed YCB objects. The results reveal a total accuracy of 81.9%, a significant improvement compared to the 34.7% achieved by the classifier trained solely on simulated images. This demonstrates the effectiveness of our approach. We further validate our approach using the classifier on a 6D object pose estimation task from tactile data.

Stefano Berti, Giulia Pasquale, Lorenzo Natale

Few-Shot Action Recognition (FS-AR) has shown promising results but is often limited by a closed-set assumption that fails in real-world open-set scenarios. While Few-Shot Open-Set (FSOS) recognition is well-established for images, its extension to spatio-temporal video data remains underexplored. To address this, we propose an architectural extension based on a Feature-Residual Discriminator (FR-Disc), adapting previous work on skeletal data to the more complex video domain. Extensive experiments on five datasets demonstrate that while common open-set techniques provide only marginal gains, our FR-Disc significantly enhances unknown rejection capabilities without compromising closed-set accuracy, setting a new state-of-the-art for FSOS-AR. The project website, code, and benchmark are available at: https://hsp-iit.github.io/fsosar/.

Federico Ceola, Lorenzo Natale, Niko Sünderhauf et al.

Instructing a robot to complete an everyday task within our homes has been a long-standing challenge for robotics. While recent progress in language-conditioned imitation learning and offline reinforcement learning has demonstrated impressive performance across a wide range of tasks, they are typically limited to short-horizon tasks -- not reflective of those a home robot would be expected to complete. While existing architectures have the potential to learn these desired behaviours, the lack of the necessary long-horizon, multi-step datasets for real robotic systems poses a significant challenge. To this end, we present the Long-Horizon Manipulation (LHManip) dataset comprising 200 episodes, demonstrating 20 different manipulation tasks via real robot teleoperation. The tasks entail multiple sub-tasks, including grasping, pushing, stacking and throwing objects in highly cluttered environments. Each task is paired with a natural language instruction and multi-camera viewpoints for point-cloud or NeRF reconstruction. In total, the dataset comprises 176,278 observation-action pairs which form part of the Open X-Embodiment dataset. The full LHManip dataset is made publicly available at https://github.com/fedeceola/LHManip.

Federico Vasile, Ri-Zhao Qiu, Lorenzo Natale et al.

System identification involving the geometry, appearance, and physical properties from video observations is a challenging task with applications in robotics and graphics. Recent approaches have relied on fully differentiable Material Point Method (MPM) and rendering for simultaneous optimization of these properties. However, they are limited to simplified object-environment interactions with planar colliders and fail in more challenging scenarios where objects collide with non-planar surfaces. We propose AS-DiffMPM, a differentiable MPM framework that enables physical property estimation with arbitrarily shaped colliders. Our approach extends existing methods by incorporating a differentiable collision handling mechanism, allowing the target object to interact with complex rigid bodies while maintaining end-to-end optimization. We show AS-DiffMPM can be easily interfaced with various novel view synthesis methods as a framework for system identification from visual observations.

Gabriele Mario Caddeo, Pasquale Marra, Lorenzo Natale

We propose a multimodal, physically grounded approach for metric-scale amodal object reconstruction and pose estimation under severe hand occlusion. Unlike prior occlusion-aware 3D generation methods that rely only on vision, we leverage physical interaction signals: proprioception provides the posed hand geometry, and multi-contact touch constrains where the object surface must lie, reducing ambiguity in occluded regions. We represent object structure as a pose-aware, camera-aligned signed distance field (SDF) and learn a compact latent space with a Structure-VAE. In this latent space, we train a conditional flow-matching diffusion model, pretraining on vision-only images and finetuning on occluded manipulation scenes while conditioning on visible RGB evidence, occluder/visibility masks, the hand latent representation, and tactile information. Crucially, we incorporate physics-based objectives and differentiable decoder-guidance during finetuning and inference to reduce hand--object interpenetration and to align the reconstructed surface with contact observations. Because our method produces a metric, physically consistent structure estimate, it integrates naturally into existing two-stage reconstruction pipelines, where a downstream module refines geometry and predicts appearance. Experiments in simulation show that adding proprioception and touch substantially improves completion under occlusion and yields physically plausible reconstructions at correct real-world scale compared to vision-only baselines; we further validate transfer by deploying the model on a real humanoid robot with an end-effector different from those used during training.

Pasquale Marra, Gabriele M. Caddeo, Ugo Pattacini et al.

In this paper, we address force-aware control and force distribution in robotic platforms with multi-fingered hands. Given a target goal and force estimates from tactile sensors, we design a controller that adapts the motion of the torso, arm, wrist, and fingers, redistributing forces to maintain stable contact with objects of varying mass distribution or unstable contacts. To estimate forces, we collect a dataset of tactile signals and ground-truth force measurements using five Xela magnetic sensors interacting with indenters, and train force estimators. We then introduce a model-based control scheme that minimizes the distance between the Center of Pressure (CoP) and the centroid of the fingertips contact polygon. Since our method relies on estimated forces rather than raw tactile signals, it has the potential to be applied to any sensor capable of force estimation. We validate our framework on a balancing task with five objects, achieving a $82.7\%$ success rate, and further evaluate it in multi-object scenarios, achieving $80\%$ accuracy. Code and data can be found here https://github.com/hsp-iit/multifingered-force-aware-control.

Michele Colledanchise, Lorenzo Natale

This paper addresses the concurrency issues affecting Behavior Trees (BTs), a popular tool to model the behaviors of autonomous agents in the video game and the robotics industry. BT designers can easily build complex behaviors composing simpler ones, which represents a key advantage of BTs. The parallel composition of BTs expresses a way to combine concurrent behaviors that has high potential, since composing pre-existing BTs in parallel results easier than composing in parallel classical control architectures, as finite state machines or teleo-reactive programs. However, BT designers rarely use such composition due to the underlying concurrency problems similar to the ones faced in concurrent programming. As a result, the parallel composition, despite its potential, finds application only in the composition of simple behaviors or where the designer can guarantee the absence of conflicts by design. In this paper, we define two new BT nodes to tackle the concurrency problems in BTs and we show how to exploit them to create predictable behaviors. In addition, we introduce measures to assess execution performance and show how different design choices affect them. We validate our approach in both simulations and the real world. Simulated experiments provide statistically significant data, whereas real-world experiments show the applicability of our method on real robots. We provided an open-source implementation of the novel BT formulation and published all the source code to reproduce the numerical examples and experiments.

Michele Colledanchise, Lorenzo Natale

There is a growing interest in Behavior Trees (BTs) as a tool to describe and implement robot behaviors. BTs were devised in the video game industry and their adoption in robotics resulted in the development of ad-hoc libraries to design and execute BTs that fit complex robotics software architectures. While there is broad consensus on how BTs work, some characteristics rely on the implementation choices done in the specific software library used. In this letter, we outline practical aspects in the adoption of BTs and the solutions devised by the robotics community to fully exploit the advantages of BTs in real robots. We also overview the solutions proposed in open-source libraries used in robotics, we show how BTs fit in robotic software architecture, and we present a use case example.

Andrea Maracani, Lorenzo Rosasco, Lorenzo Natale

Deep Neural Networks have significantly impacted many computer vision tasks. However, their effectiveness diminishes when test data distribution (target domain) deviates from the one of training data (source domain). In situations where target labels are unavailable and the access to the labeled source domain is restricted due to data privacy or memory constraints, Source-Free Unsupervised Domain Adaptation (SF-UDA) has emerged as a valuable tool. Recognizing the key role of SF-UDA under these constraints, we introduce a novel approach marked by two key contributions: Few Trusted Samples Pseudo-labeling (FTSP) and Temperature Scaled Adaptive Loss (TSAL). FTSP employs a limited subset of trusted samples from the target data to construct a classifier to infer pseudo-labels for the entire domain, showing simplicity and improved accuracy. Simultaneously, TSAL, designed with a unique dual temperature scheduling, adeptly balance diversity, discriminability, and the incorporation of pseudo-labels in the unsupervised adaptation objective. Our methodology, that we name Trust And Balance (TAB) adaptation, is rigorously evaluated on standard datasets like Office31 and Office-Home, and on less common benchmarks such as ImageCLEF-DA and Adaptiope, employing both ResNet50 and ViT-Large architectures. Our results compare favorably with, and in most cases surpass, contemporary state-of-the-art techniques, underscoring the effectiveness of our methodology in the SF-UDA landscape.

Francesco Taioli, Stefano Rosa, Alberto Castellini et al.

Vision-and-Language Navigation in Continuous Environments (VLN-CE) is one of the most intuitive yet challenging embodied AI tasks. Agents are tasked to navigate towards a target goal by executing a set of low-level actions, following a series of natural language instructions. All VLN-CE methods in the literature assume that language instructions are exact. However, in practice, instructions given by humans can contain errors when describing a spatial environment due to inaccurate memory or confusion. Current VLN-CE benchmarks do not address this scenario, making the state-of-the-art methods in VLN-CE fragile in the presence of erroneous instructions from human users. For the first time, we propose a novel benchmark dataset that introduces various types of instruction errors considering potential human causes. This benchmark provides valuable insight into the robustness of VLN systems in continuous environments. We observe a noticeable performance drop (up to -25%) in Success Rate when evaluating the state-of-the-art VLN-CE methods on our benchmark. Moreover, we formally define the task of Instruction Error Detection and Localization, and establish an evaluation protocol on top of our benchmark dataset. We also propose an effective method, based on a cross-modal transformer architecture, that achieves the best performance in error detection and localization, compared to baselines. Surprisingly, our proposed method has revealed errors in the validation set of the two commonly used datasets for VLN-CE, i.e., R2R-CE and RxR-CE, demonstrating the utility of our technique in other tasks. Code and dataset available at https://intelligolabs.github.io/R2RIE-CE

Giuseppe Stracquadanio, Federico Vasile, Elisa Maiettini et al.

One of the most important research challenges in upper-limb prosthetics is enhancing the user-prosthesis communication to closely resemble the experience of a natural limb. As prosthetic devices become more complex, users often struggle to control the additional degrees of freedom. In this context, leveraging shared-autonomy principles can significantly improve the usability of these systems. In this paper, we present a novel eye-in-hand prosthetic grasping system that follows these principles. Our system initiates the approach-to-grasp action based on user's command and automatically configures the DoFs of a prosthetic hand. First, it reconstructs the 3D geometry of the target object without the need of a depth camera. Then, it tracks the hand motion during the approach-to-grasp action and finally selects a candidate grasp configuration according to user's intentions. We deploy our system on the Hannes prosthetic hand and test it on able-bodied subjects and amputees to validate its effectiveness. We compare it with a multi-DoF prosthetic control baseline and find that our method enables faster grasps, while simplifying the user experience. Code and demo videos are available online at https://hsp-iit.github.io/byogg/.

Federico Vasile, Elisa Maiettini, Giulia Pasquale et al.

Most control techniques for prosthetic grasping focus on dexterous fingers control, but overlook the wrist motion. This forces the user to perform compensatory movements with the elbow, shoulder and hip to adapt the wrist for grasping. We propose a computer vision-based system that leverages the collaboration between the user and an automatic system in a shared autonomy framework, to perform continuous control of the wrist degrees of freedom in a prosthetic arm, promoting a more natural approach-to-grasp motion. Our pipeline allows to seamlessly control the prosthetic wrist to follow the target object and finally orient it for grasping according to the user intent. We assess the effectiveness of each system component through quantitative analysis and finally deploy our method on the Hannes prosthetic arm. Code and videos: https://hsp-iit.github.io/hannes-wrist-control.

Andrea Maracani, Raffaello Camoriano, Elisa Maiettini et al.

This study provides a comprehensive benchmark framework for Source-Free Unsupervised Domain Adaptation (SF-UDA) in image classification, aiming to achieve a rigorous empirical understanding of the complex relationships between multiple key design factors in SF-UDA methods. The study empirically examines a diverse set of SF-UDA techniques, assessing their consistency across datasets, sensitivity to specific hyperparameters, and applicability across different families of backbone architectures. Moreover, it exhaustively evaluates pre-training datasets and strategies, particularly focusing on both supervised and self-supervised methods, as well as the impact of fine-tuning on the source domain. Our analysis also highlights gaps in existing benchmark practices, guiding SF-UDA research towards more effective and general approaches. It emphasizes the importance of backbone architecture and pre-training dataset selection on SF-UDA performance, serving as an essential reference and providing key insights. Lastly, we release the source code of our experimental framework. This facilitates the construction, training, and testing of SF-UDA methods, enabling systematic large-scale experimental analysis and supporting further research efforts in this field.

Zhichao Li, Arren Glover, Chiara Bartolozzi et al.

Tracking the position and orientation of objects in space (i.e., in 6-DoF) in real time is a fundamental problem in robotics for environment interaction. It becomes more challenging when objects move at high-speed due to frame rate limitations in conventional cameras and motion blur. Event cameras are characterized by high temporal resolution, low latency and high dynamic range, that can potentially overcome the impacts of motion blur. Traditional RGB cameras provide rich visual information that is more suitable for the challenging task of single-shot object pose estimation. In this work, we propose using event-based optical flow combined with an RGB based global object pose estimator for 6-DoF pose tracking of objects at high-speed, exploiting the core advantages of both types of vision sensors. Specifically, we propose an event-based optical flow algorithm for object motion measurement to implement an object 6-DoF velocity tracker. By integrating the tracked object 6-DoF velocity with low frequency estimated pose from the global pose estimator, the method can track pose when objects move at high-speed. The proposed algorithm is tested and validated on both synthetic and real world data, demonstrating its effectiveness, especially in high-speed motion scenarios.

Carlo Alessi, Federico Vasile, Federico Ceola et al.

Recent advancements in control of prosthetic hands have focused on increasing autonomy through the use of cameras and other sensory inputs. These systems aim to reduce the cognitive load on the user by automatically controlling certain degrees of freedom. In robotics, imitation learning has emerged as a promising approach for learning grasping and complex manipulation tasks while simplifying data collection. Its application to the control of prosthetic hands remains, however, largely unexplored. Bridging this gap could enhance dexterity restoration and enable prosthetic devices to operate in more unconstrained scenarios, where tasks are learned from demonstrations rather than relying on manually annotated sequences. To this end, we present HannesImitationPolicy, an imitation learning-based method to control the Hannes prosthetic hand, enabling object grasping in unstructured environments. Moreover, we introduce the HannesImitationDataset comprising grasping demonstrations in table, shelf, and human-to-prosthesis handover scenarios. We leverage such data to train a single diffusion policy and deploy it on the prosthetic hand to predict the wrist orientation and hand closure for grasping. Experimental evaluation demonstrates successful grasps across diverse objects and conditions. Finally, we show that the policy outperforms a segmentation-based visual servo controller in unstructured scenarios. Additional material is provided on our project page: https://hsp-iit.github.io/HannesImitation

Tommaso Galliena, Tommaso Apicella, Stefano Rosa et al.

We present a self-supervised method to improve an agent's abilities in describing arbitrary objects while actively exploring a generic environment. This is a challenging problem, as current models struggle to obtain coherent image captions due to different camera viewpoints and clutter. We propose a three-phase framework to fine-tune existing captioning models that enhances caption accuracy and consistency across views via a consensus mechanism. First, an agent explores the environment, collecting noisy image-caption pairs. Then, a consistent pseudo-caption for each object instance is distilled via consensus using a large language model. Finally, these pseudo-captions are used to fine-tune an off-the-shelf captioning model, with the addition of contrastive learning. We analyse the performance of the combination of captioning models, exploration policies, pseudo-labeling methods, and fine-tuning strategies, on our manually labeled test set. Results show that a policy can be trained to mine samples with higher disagreement compared to classical baselines. Our pseudo-captioning method, in combination with all policies, has a higher semantic similarity compared to other existing methods, and fine-tuning improves caption accuracy and consistency by a significant margin. Code and test set annotations available at https://hsp-iit.github.io/embodied-captioning/

Francesco Taioli, Stefano Rosa, Alberto Castellini et al.

In the Vision-and-Language Navigation in Continuous Environments (VLN-CE) task, the human user guides an autonomous agent to reach a target goal via a series of low-level actions following a textual instruction in natural language. However, most existing methods do not address the likely case where users may make mistakes when providing such instruction (e.g. "turn left" instead of "turn right"). In this work, we address a novel task of Interactive VLN in Continuous Environments (IVLN-CE), which allows the agent to interact with the user during the VLN-CE navigation to verify any doubts regarding the instruction errors. We propose an Interactive Instruction Error Detector and Localizer (I2EDL) that triggers the user-agent interaction upon the detection of instruction errors during the navigation. We leverage a pre-trained module to detect instruction errors and pinpoint them in the instruction by cross-referencing the textual input and past observations. In such way, the agent is able to query the user for a timely correction, without demanding the user's cognitive load, as we locate the probable errors to a precise part of the instruction. We evaluate the proposed I2EDL on a dataset of instructions containing errors, and further devise a novel metric, the Success weighted by Interaction Number (SIN), to reflect both the navigation performance and the interaction effectiveness. We show how the proposed method can ask focused requests for corrections to the user, which in turn increases the navigation success, while minimizing the interactions.

Carmela Calabrese, Stefano Berti, Giulia Pasquale et al.

Addressing multi-label action recognition in videos represents a significant challenge for robotic applications in dynamic environments, especially when the robot is required to cooperate with humans in tasks that involve objects. Existing methods still struggle to recognize unseen actions or require extensive training data. To overcome these problems, we propose Dual-VCLIP, a unified approach for zero-shot multi-label action recognition. Dual-VCLIP enhances VCLIP, a zero-shot action recognition method, with the DualCoOp method for multi-label image classification. The strength of our method is that at training time it only learns two prompts, and it is therefore much simpler than other methods. We validate our method on the Charades dataset that includes a majority of object-based actions, demonstrating that -- despite its simplicity -- our method performs favorably with respect to existing methods on the complete dataset, and promising performance when tested on unseen actions. Our contribution emphasizes the impact of verb-object class-splits during robots' training for new cooperative tasks, highlighting the influence on the performance and giving insights into mitigating biases.

Nicola A. Piga, Yuriy Onyshchuk, Giulia Pasquale et al.

6D object pose tracking has been extensively studied in the robotics and computer vision communities. The most promising solutions, leveraging on deep neural networks and/or filtering and optimization, exhibit notable performance on standard benchmarks. However, to our best knowledge, these have not been tested thoroughly against fast object motions. Tracking performance in this scenario degrades significantly, especially for methods that do not achieve real-time performance and introduce non negligible delays. In this work, we introduce ROFT, a Kalman filtering approach for 6D object pose and velocity tracking from a stream of RGB-D images. By leveraging real-time optical flow, ROFT synchronizes delayed outputs of low frame rate Convolutional Neural Networks for instance segmentation and 6D object pose estimation with the RGB-D input stream to achieve fast and precise 6D object pose and velocity tracking. We test our method on a newly introduced photorealistic dataset, Fast-YCB, which comprises fast moving objects from the YCB model set, and on the dataset for object and hand pose estimation HO-3D. Results demonstrate that our approach outperforms state-of-the-art methods for 6D object pose tracking, while also providing 6D object velocity tracking. A video showing the experiments is provided as supplementary material.

Evgenii Safronov, Nicola Piga, Michele Colledanchise et al.

Recent visual pose estimation and tracking solutions provide notable results on popular datasets such as T-LESS and YCB. However, in the real world, we can find ambiguous objects that do not allow exact classification and detection from a single view. In this work, we propose a framework that, given a single view of an object, provides the coordinates of a next viewpoint to discriminate the object against similar ones, if any, and eliminates ambiguities. We also describe a complete pipeline from a real object's scans to the viewpoint selection and classification. We validate our approach with a Franka Emika Panda robot and common household objects featured with ambiguities. We released the source code to reproduce our experiments.

Elisa Maiettini, Vadim Tikhanoff, Lorenzo Natale

Reliable perception and efficient adaptation to novel conditions are priority skills for humanoids that function in dynamic environments. The vast advancements in latest computer vision research, brought by deep learning methods, are appealing for the robotics community. However, their adoption in applied domains is not straightforward since adapting them to new tasks is strongly demanding in terms of annotated data and optimization time. Nevertheless, robotic platforms, and especially humanoids, present opportunities (such as additional sensors and the chance to explore the environment) that can be exploited to overcome these issues. In this paper, we present a pipeline for efficiently training an object detection system on a humanoid robot. The proposed system allows to iteratively adapt an object detection model to novel scenarios, by exploiting: (i) a teacher-learner pipeline, (ii) weakly supervised learning techniques to reduce the human labeling effort and (iii) an on-line learning approach for fast model re-training. We use the R1 humanoid robot for both testing the proposed pipeline in a real-time application and acquire sequences of images to benchmark the method. We made the code of the application publicly available.

Michele Colledanchise, Giuseppe Cicala, Daniele E. Domenichelli et al.

In this paper, we present a toolchain to design, execute, and verify robot behaviors. The toolchain follows the guidelines defined by the EU H2020 project RobMoSys and encodes the robot deliberation as a Behavior Tree (BT), a directed tree where the internal nodes model behavior composition and leaf nodes model action or measurement operations. Such leaf nodes take the form of a statechart (SC), which runs in separate threads, whose states perform basic arithmetic operations and send commands to the robot. The toolchain provides the ability to define a runtime monitor for a given system specification that warns the user whenever a given specification is violated. We validated the toolchain in a simulated experiment that we made reproducible in an OS-virtualization environment.

Michele Colledanchise, Giuseppe Cicala, Daniele E. Domenichelli et al.

In this paper, we enable automated property verification of deliberative components in robot control architectures. We focus on formalizing the execution context of Behavior Trees (BTs) to provide a scalable, yet formally grounded, methodology to enable runtime verification and prevent unexpected robot behaviors. To this end, we consider a message-passing model that accommodates both synchronous and asynchronous composition of parallel components, in which BTs and other components execute and interact according to the communication patterns commonly adopted in robotic software architectures. We introduce a formal property specification language to encode requirements and build runtime monitors. We performed a set of experiments, both on simulations and on the real robot, demonstrating the feasibility of our approach in a realistic application and its integration in a typical robot software architecture. We also provide an OS-level virtualization environment to reproduce the experiments in the simulated scenario.

Lorenzo Natale, Giorgio Cannata

Research on tactile sensing has been progressing at constant pace. In robotics, tactile sensing is typically studied in the context of object grasping and manipulation. In this domain, the development of robust, multi-modal, tactile sensors for robotic hands has supported the study of novel algorithms for in-hand object manipulation, material classification and object perception. In the field of humanoid robotics, research has focused on solving the challenges that allow developing systems of tactile sensors that can cover large areas of the robot body, and can integrate different types of transducers to measure pressure at various frequency bands, acceleration and temperature. The availability of such systems has extended the application of tactile sensing to whole-body control, autonomous calibration, self-perception and human-robot interaction. The goal of this Chapter is to provide an overview of the technologies for tactile sensing, with particular emphasis on the systems that have been deployed on humanoid robots. We describe the skills that have been implemented with the adoption of these technologies and discuss the main challenges that remain to be addressed.

Lorenzo Natale, Chiara Bartolozzi, Francesco Nori et al.

In this chapter we describe the history and evolution of the iCub humanoid platform. We start by describing the first version as it was designed during the RobotCub EU project and illustrate how it evolved to become the platform that is adopted by more than 30 laboratories world wide. We complete the chapter by illustrating some of the research activities that are currently carried out on the iCub robot, i.e. visual perception, event driven sensing and dynamic control. We conclude the Chapter with a discussion of the lessons we learned and a preview of the upcoming next release of the robot, iCub 3.0.

Prajval Kumar Murali, Ines Sorrentino, Angelo Rendiniello et al.

Hand-eye calibration of laser profile sensors is the process of extracting the homogeneous transformation between the laser profile sensor frame and the end-effector frame of a robot in order to express the data extracted by the sensor in the robot's global coordinate system. For laser profile scanners this is a challenging procedure, as they provide data only in two dimensions and state-of-the-art calibration procedures require the use of specialised calibration targets. This paper presents a novel method to extract the translation-part of the hand-eye calibration matrix with rotation-part known a priori in a target-agnostic way. Our methodology is applicable to any 2D image or 3D object as a calibration target and can also be performed in situ in the final application. The method is experimentally validated on a real robot-sensor setup with 2D and 3D targets.

Elisa Maiettini, Andrea Maracani, Raffaello Camoriano et al.

Deep Learning (DL) based methods for object detection achieve remarkable performance at the cost of computationally expensive training and extensive data labeling. Robots embodiment can be exploited to mitigate this burden by acquiring automatically annotated training data via a natural interaction with a human showing the object of interest, handheld. However, learning solely from this data may introduce biases (the so-called domain shift), and prevents adaptation to novel tasks. While Weakly-supervised Learning (WSL) offers a well-established set of techniques to cope with these problems in general-purpose Computer Vision, its adoption in challenging robotic domains is still at a preliminary stage. In this work, we target the scenario of a robot trained in a teacher-learner setting to detect handheld objects. The aim is to improve detection performance in different settings by letting the robot explore the environment with a limited human labeling budget. We compare several techniques for WSL in detection pipelines to reduce model re-training costs without compromising accuracy, proposing solutions which target the considered robotic scenario. We show that the robot can improve adaptation to novel domains, either by interacting with a human teacher (Active Learning) or with an autonomous supervision (Semi-supervised Learning). We integrate our strategies into an on-line detection method, achieving efficient model update capabilities with few labels. We experimentally benchmark our method on challenging robotic object detection tasks under domain shift.

Federico Ceola, Elisa Maiettini, Giulia Pasquale et al.

Object segmentation is a key component in the visual system of a robot that performs tasks like grasping and object manipulation, especially in presence of occlusions. Like many other computer vision tasks, the adoption of deep architectures has made available algorithms that perform this task with remarkable performance. However, adoption of such algorithms in robotics is hampered by the fact that training requires large amount of computing time and it cannot be performed on-line. In this work, we propose a novel architecture for object segmentation, that overcomes this problem and provides comparable performance in a fraction of the time required by the state-of-the-art methods. Our approach is based on a pre-trained Mask R-CNN, in which various layers have been replaced with a set of classifiers and regressors that are re-trained for a new task. We employ an efficient Kernel-based method that allows for fast training on large scale problems. Our approach is validated on the YCB-Video dataset which is widely adopted in the computer vision and robotics community, demonstrating that we can achieve and even surpass performance of the state-of-the-art, with a significant reduction (${\sim}6\times$) of the training time. The code to reproduce the experiments is publicly available on GitHub.

Federico Ceola, Elisa Maiettini, Giulia Pasquale et al.

Object detection is a fundamental task for robots to operate in unstructured environments. Today, there are several deep learning algorithms that solve this task with remarkable performance. Unfortunately, training such systems requires several hours of GPU time. For robots, to successfully adapt to changes in the environment or learning new objects, it is also important that object detectors can be re-trained in a short amount of time. A recent method [1] proposes an architecture that leverages on the powerful representation of deep learning descriptors, while permitting fast adaptation time. Leveraging on the natural decomposition of the task in (i) regions candidate generation, (ii) feature extraction and (iii) regions classification, this method performs fast adaptation of the detector, by only re-training the classification layer. This shortens training time while maintaining state-of-the-art performance. In this paper, we firstly demonstrate that a further boost in accuracy can be obtained by adapting, in addition, the regions candidate generation on the task at hand. Secondly, we extend the object detection system presented in [1] with the proposed fast learning approach, showing experimental evidence on the improvement provided in terms of speed and accuracy on two different robotics datasets. The code to reproduce the experiments is publicly available on GitHub.

Evgenii Safronov, Michele Colledanchise, Lorenzo Natale

In this paper, we propose Belief Behavior Trees (BBTs), an extension to Behavior Trees (BTs) that allows to automatically create a policy that controls a robot in partially observable environments. We extend the semantic of BTs to account for the uncertainty that affects both the conditions and action nodes of the BT. The tree gets synthesized following a planning strategy for BTs proposed recently: from a set of goal conditions we iteratively select a goal and find the action, or in general the subtree, that satisfies it. Such action may have preconditions that do not hold. For those preconditions, we find an action or subtree in the same fashion. We extend this approach by including, in the planner, actions that have the purpose to reduce the uncertainty that affects the value of a condition node in the BT (for example, turning on the lights to have better lighting conditions). We demonstrate that BBTs allows task planning with non-deterministic outcomes for actions. We provide experimental validation of our approach in a real robotic scenario and - for sake of reproducibility - in a simulated one.

Evgenii Safronov, Michele Colledanchise, Lorenzo Natale

Task planning in a probabilistic belief state domains allows generating complex and robust execution policies in those domains affected by state uncertainty. The performance of a task planner relies on the belief state representation. However, current belief state representation becomes easily intractable as the number of variables and execution time grows. To address this problem, we developed a novel belief state representation based on cartesian product and union operations over belief substates. These two operations and single variable assignment nodes form And-Or directed acyclic graph of Belief State (AOBS). We show how to apply actions with probabilistic outcomes and measure the probability of conditions holding over belief state. We evaluated AOBS performance in simulated forward state space exploration. We compared the size of AOBS with the size of Binary Decision Diagrams (BDD) that were previously used to represent belief state. We show that AOBS representation is not only much more compact than a full belief state but it also scales better than BDD for most of the cases.

Michele Colledanchise, Damiano Malafronte, Lorenzo Natale

In this paper, we outline an interleaved acting and planning technique to rapidly reduce the uncertainty of the estimated robot's pose by perceiving relevant information from the environment, as recognizing an object or asking someone for a direction. Generally, existing localization approaches rely on low-level geometric features such as points, lines, and planes, while these approaches provide the desired accuracy, they may require time to converge, especially with incorrect initial guesses. In our approach, a task planner computes a sequence of action and perception tasks to actively obtain relevant information from the robot's perception system. We validate our approach in large state spaces, to show how the approach scales, and in real environments, to show the applicability of our method on real robots. We prove that our approach is sound, probabilistically complete, and tractable in practical cases.

Fabrizio Bottarel, Giulia Vezzani, Ugo Pattacini et al.

The use of benchmarks is a widespread and scientifically meaningful practice to validate performance of different approaches to the same task. In the context of robot grasping the use of common object sets has emerged in recent years, however no dominant protocols and metrics to test grasping pipelines have taken root yet. In this paper, we present version 1.0 of GRASPA, a benchmark to test effectiveness of grasping pipelines on physical robot setups. This approach tackles the complexity of such pipelines by proposing different metrics that account for the features and limits of the test platform. As an example application, we deploy GRASPA on the iCub humanoid robot and use it to benchmark our grasping pipeline. As closing remarks, we discuss how the GRASPA indicators we obtained as outcome can provide insight into how different steps of the pipeline affect the overall grasping performance.

Michele Colledanchise, Lorenzo Natale

Behavior Trees (BTs) are becoming a popular tool to model the behaviors of autonomous agents in the computer game and the robotics industry. One of the key advantages of BTs lies in their composability, where complex behaviors can be built by composing simpler ones. The parallel composition is the one with the highest potential since the complexity of composing pre-existing behaviors in parallel is much lower than the one needed using classical control architectures as finite state machines. However, the parallel composition is rarely used due to the underlying concurrency problems that are similar to the ones faced in concurrent programming. In this paper, we define two synchronization techniques to tackle the concurrency problems in BTs compositions and we show how to exploit them to improve behavior predictability. Also, we introduce measures to assess execution performance, and we show how design choices can affect them. To illustrate the proposed framework, we provide a set of experiments using the R1 robot and we gather statistically-significant data.

Giulia Vezzani, Abhishek Gupta, Lorenzo Natale et al.

Exploration is an extremely challenging problem in reinforcement learning, especially in high dimensional state and action spaces and when only sparse rewards are available. Effective representations can indicate which components of the state are task relevant and thus reduce the dimensionality of the space to explore. In this work, we take a representation learning viewpoint on exploration, utilizing prior experience to learn effective latent representations, which can subsequently indicate which regions to explore. Prior experience on separate but related tasks help learn representations of the state which are effective at predicting instantaneous rewards. These learned representations can then be used with an entropy-based exploration method to effectively perform exploration in high dimensional spaces by effectively lowering the dimensionality of the search space. We show the benefits of this representation for meta-exploration in a simulated object pushing environment.

Michele Colledanchise, Lorenzo Natale

Behavior Trees (BTs) have become a popular framework for designing controllers of autonomous agents in the computer game and in the robotics industry. One of the key advantages of BTs lies in their modularity, where independent modules can be composed to create more complex ones. In the classical formulation of BTs, modules can be composed using one of the three operators: Sequence, Fallback, and Parallel. The Parallel operator is rarely used despite its strong potential against other control architectures as Finite State Machines. This is due to the fact that concurrent actions may lead to unexpected problems similar to the ones experienced in concurrent programming. In this paper, we introduce Concurrent BTs (CBTs) as a generalization of BTs in which we introduce the notions of progress and resource usage. We show how CBTs allow safe concurrent executions of actions and we analyze the approach from a mathematical standpoint. To illustrate the use of CBTs, we provide a set of use cases in robotics scenarios.

Elisa Maiettini, Giulia Pasquale, Lorenzo Rosasco et al.

Latest deep learning methods for object detection provide remarkable performance, but have limits when used in robotic applications. One of the most relevant issues is the long training time, which is due to the large size and imbalance of the associated training sets, characterized by few positive and a large number of negative examples (i.e. background). Proposed approaches are based on end-to-end learning by back-propagation [22] or kernel methods trained with Hard Negatives Mining on top of deep features [8]. These solutions are effective, but prohibitively slow for on-line applications. In this paper we propose a novel pipeline for object detection that overcomes this problem and provides comparable performance, with a 60x training speedup. Our pipeline combines (i) the Region Proposal Network and the deep feature extractor from [22] to efficiently select candidate RoIs and encode them into powerful representations, with (ii) the FALKON [23] algorithm, a novel kernel-based method that allows fast training on large scale problems (millions of points). We address the size and imbalance of training data by exploiting the stochastic subsampling intrinsic into the method and a novel, fast, bootstrapping approach. We assess the effectiveness of the approach on a standard Computer Vision dataset (PASCAL VOC 2007 [5]) and demonstrate its applicability to a real robotic scenario with the iCubWorld Transformations [18] dataset.

Claudio Fantacci, Giulia Vezzani, Ugo Pattacini et al.

To precisely reach for an object with a humanoid robot, it is of central importance to have good knowledge of both end-effector, object pose and shape. In this work we propose a framework for markerless visual servoing on unknown objects, which is divided in four main parts: I) a least-squares minimization problem is formulated to find the volume of the object graspable by the robot's hand using its stereo vision; II) a recursive Bayesian filtering technique, based on Sequential Monte Carlo (SMC) filtering, estimates the 6D pose (position and orientation) of the robot's end-effector without the use of markers; III) a nonlinear constrained optimization problem is formulated to compute the desired graspable pose about the object; IV) an image-based visual servo control commands the robot's end-effector toward the desired pose. We demonstrate effectiveness and robustness of our approach with extensive experiments on the iCub humanoid robot platform, achieving real-time computation, smooth trajectories and sub-pixel precisions.

Giulia Pasquale, Carlo Ciliberto, Francesca Odone et al.

We report on an extensive study of the benefits and limitations of current deep learning approaches to object recognition in robot vision scenarios, introducing a novel dataset used for our investigation. To avoid the biases in currently available datasets, we consider a natural human-robot interaction setting to design a data-acquisition protocol for visual object recognition on the iCub humanoid robot. Analyzing the performance of off-the-shelf models trained off-line on large-scale image retrieval datasets, we show the necessity for knowledge transfer. We evaluate different ways in which this last step can be done, and identify the major bottlenecks affecting robotic scenarios. By studying both object categorization and identification problems, we highlight key differences between object recognition in robotics applications and in image retrieval tasks, for which the considered deep learning approaches have been originally designed. In a nutshell, our results confirm the remarkable improvements yield by deep learning in this setting, while pointing to specific open challenges that need be addressed for seamless deployment in robotics.