Publications

Background: Foot drop syndrome (FDS), characterized by severe weakness and atrophy of the dorsiflexion muscles of the feet, is commonly found in patients with severe acquired brain injury (ABI). If the syndrome is unilateral, the cause is often a peroneal neuropathy (PN), due to compression of the nervous trunk on the neck of the fibula at the knee level; less frequently, the cause is a previous or concomitant lumbar radiculopathy. Bilateral syndromes are caused by polyneuropathies and myopathies. Central causes, due to brain or spinal injury, mimic this syndrome but are usually accompanied by other symptoms, such as spasticity. Critical illness polyneuropathy (CIP) and myopathy (CIM), isolated or in combination (critical illness polyneuromyopathy, CIPNM), have been shown to constitute an important cause of FDS in patients with ABI. Assessing the causes of FDS in the intensive rehabilitation unit (IRU) has several limitations, which include the complexity of the electrophysiological tests, limited availability of neurophysiology consultants, and the severe disturbance in consciousness and lack of cooperation from patients. Objectives: We sought to propose a simplified electrophysiological screening that identifies FDS causes, particularly PN and CIPNM, to help clinicians to recognize the significant clinical predictors of poor outcomes in severe ABI at admission to IRU. Methods: This prospective, single-center study included 20 severe ABI patients with FDS (11 females/9 males, mean age 55.10 + 16.26; CRS-R= 11.90 + 6.32; LCF: 3.30 + 1.30; DRS: 21.45 + 3.33), with prolonged rehabilitation treatment (≥2 months). We applied direct tibialis anterior muscle stimulation (DMS) associated with peroneal nerve motor conduction evaluation, across the fibular head (NCS), to identify CIP and/or CIM and to exclude demyelinating or compressive unilateral PN. Results: At admission to IRU, simplified electrophysiological screening reported four unilateral PN, four CIP and six CIM with a CIPNM overall prevalence estimate of about 50%. After 2 months, the CIPNM group showed significantly poorer outcomes compared to other ABI patients without CIPNM, as demonstrated by the lower probability of achieving endotracheal-tube weaning (20% versus 90%) and lower CRS-R and DRS scores. Due to the subacute rehabilitation setting of our study, it was not possible to evaluate the motor results of recovery of the standing position, functional walking and balance, impaired by the presence of unilateral PN. Conclusions: The implementation of the proposed simplified electrophysiological screening may enable the early identification of unilateral PN or CIPNM in severe ABI patients, thereby contributing to better functional prognosis in rehabilitative settings.

The ability to discern human intentions from brain signals has opened the possibility of leveraging Brain-Machine Interfaces (BMIs) for the control of robotic devices, especially benefiting individuals with severe motor disabilities. In this work, we present a novel approach for navigating a semiautonomous wheelchair towards targets generated by a BMI, all while ensuring collision avoidance. Our approach employs Nonlinear Model Predictive Control (NMPC) for real-time trajectory generation in unknown and dynamic environments. The empirical results obtained from real-world experiments clearly demonstrate the advancements of our solution over current state-of-the-art techniques. Our implementation is proven to outperform well-established methods in terms of both smoothness and alignment with the user’s intended behavior.

Recent studies in the domain of invasive brain-computer interfaces (BCIs) have revealed that neural activity recorded during the observation of robotic movements in a reach-and-grasp task carries information that can be utilized to improve the active online decoding of motor intention. In the non-invasive domain, the spectral characteristics of human brain activity during the observation of robotic movements has been widely investigated. However, focusing only on the frequency components of electroencephalography (EEG) for motor control decoding is a poorly suitable strategy due to its scarce temporal resolution. Following a different approach, we explored temporal features of EEG filtered in the delta band (Low-Frequency EEG, or LF-EEG) for the continuous decoding of control-oriented kinematic trajectories. We designed an experimental paradigm aimed at investigating how the observation of center-out target-oriented reaching movements executed by a robotic arm in the 2D plane is encoded in low-frequency EEG signals. By employing machine learning algorithms and novel approaches, we were able to continuously decode the LF-EEG into movement trajectories, achieving performance significantly above chance-level. This confirms that low-frequency neural activity measured non-invasively during a movement observation task contains adequate amounts of movement-related information for BCI applications.

This work presents and evaluates a method for reducing the number of hyper-parameters in the continuous control system used by a 2-class motor imagery (MI) brain-machine interface (BMI). The work focuses on two parameters (ω and ψ) used within a dynamical control system that considers the nature and temporal evolution of the BMI decoder output and that it has been already validated in the past.To identify the optimal values for the parameters, we analysed a dataset of 12 subjects performing 2-class MI tasks. For each subject, we defined a new metric to investigate the existence of a relationship between the hyper-parameters. The study reveals a quadratic relationship with coefficient of determination (R2) of 81.67%.Finally, the established relationship was evaluated through an closed-loop experiment involving three healthy subjects. Results demonstrated the potential use of the discovered quadratic relationship to reduce the number of parameters for the dynamical control system and, thus, to simplify the BMI operations.

Deploying service robots in our daily life, whether in restaurants, warehouses or hospitals, calls for the need to reason on the interactions happening in dense and dynamic scenes. In this paper, we present and benchmark three new approaches to model and predict multi-agent interactions in dense scenes, including the use of an intuitive qualitative representation. The proposed solutions take into account static and dynamic context to predict individual interactions. They exploit an input- and a temporal-attention mechanism, and are tested on medium and long-term time horizons. The first two approaches integrate different relations from the so-called Qualitative Trajectory Calculus (QTC) within a stateof-the-art deep neural network to create a symbol-driven neural architecture for predicting spatial interactions. The third approach implements a purely data-driven network for motion prediction, the output of which is post-processed to predict QTC spatial interactions. Experimental results on a popular robot dataset of challenging crowded scenarios show that the purely data-driven prediction approach generally outperforms the other two. The three approaches were further evaluated on a different but related human scenarios to assess their generalisation capability.

Reconstructing accurate causal models of dynamic systems from time-series of sensor data is a key problem in many real-world scenarios. In this paper, we present an overview based on our experience about practical challenges that the causal analysis encounters when applied to autonomous robots and how Continual Learning (CL) could help to overcome them. We propose a possible way to leverage the CL paradigm to make causal discovery feasible for robotics applications where the computational resources are limited, while at the same time exploiting the robot as an active agent that helps to increase the quality of the reconstructed causal models.

Identifying the main features and learning the causal relationships of a dynamic system from time-series of sensor data are key problems in many real-world robot applications. In this paper, we propose an extension of a state-of-the-art causal discovery method, PCMCI, embedding an additional feature-selection module based on transfer entropy. Starting from a prefixed set of variables, the new algorithm reconstructs the causal model of the observed system by considering only its main features and neglecting those deemed unnecessary for understanding the evolution of the system. We first validate the method on a toy problem and on synthetic data of brain network, for which the ground-truth models are available, and then on a real-world robotics scenario using a large-scale time-series dataset of human trajectories. The experiments demonstrate that our solution outperforms the previous state-of-the-art technique in terms of accuracy and computational efficiency, allowing better and faster causal discovery of meaningful models from robot sensor data.

Reasoning on the context of human beings is crucial for many real-world applications especially for those deploying autonomous systems (e.g. robots). In this paper, we present a new approach for context reasoning to further advance the field of human motion prediction. We therefore propose a neuro-symbolic approach for human motion prediction (NeuroSyM), which weights differently the interactions in the neighbourhood by leveraging an intuitive technique for spatial representation called Qualitative Trajectory Calculus (QTC). The proposed approach is experimentally tested on medium and long term time horizons using two architectures from the state of art, one of which is a baseline for human motion prediction and the other is a baseline for generic multivariate time-series prediction. Six datasets of challenging crowded scenarios, collected from both fixed and mobile cameras, were used for testing. Experimental results show that the NeuroSyM approach outperforms in most cases the baseline architectures in terms of prediction accuracy.

This work examines the viability of self-supervised learning approaches in the field of agri-robotics, specifically focusing on the segmentation of densely packed potato tubers in storage. The work assesses the impact of both the quantity and quality of data on self-supervised training, employing a limited set of both annotated and unannotated data. Mask R-CNN with a ResNet50 backbone is used for instance segmentation to evaluate self-supervised training performance. The results indicate that the self-supervised methods employed have a modest yet beneficial impact on the downstream task. A simpler approach yields more effective results with a larger dataset, whereas a more intricate method shows superior performance with a refined, smaller self-supervised dataset.

Person detection applications based on computer vision techniques often rely on complex Convolutional Neural Networks that require powerful hardware in order achieve good runtime performance. The work of this paper has been developed with the aim of implementing a safety system, based on computer vision algorithms, able to detect people in working environments using an embedded device. Possible applications for such safety systems include remote site monitoring and autonomous mobile robots in warehouses and industrial premises. Similar studies already exist in the literature, but they mostly rely on systems like NVidia Jetson that, with a Cuda enabled GPU, are able to provide satisfactory results. This, however, comes with a significant downside as such devices are usually expensive and require significant power consumption. The current paper instead is going to consider various implementations of computer vision-based person detection on two power-efficient and inexpensive devices, namely Raspberry Pi 3 and 4. In order to do so, some solutions based on off-the-shelf algorithms are first explored by reporting experimental results based on relevant performance metrics. Then, the paper presents a newly-created custom architecture, called eYOLO, that tries to solve some limitations of the previous systems. The experimental evaluation demonstrates the good performance of the proposed approach and suggests ways for further improvement.