Publications

The ActiVis project’s aim is to build a mobile guidance aid to help people with limited vision find objects in an unknown environment. This system uses bone-conduction headphones to transmit audio signals to the user and requires an effective non-visual interface. To this end, we propose a new audio-based interface that uses a spatialised signal to convey a target’s position on the horizontal plane. The vertical position on the median plan is given by adjusting the tone’s pitch to overcome the audio localisation limitations of bone-conduction headphones. This interface is validated through a set of experiments with blindfolded and visually impaired participants.

The development of autonomous robots for agriculture depends on a successful approach to recognize user needs as well as datasets reflecting the characteristics of the domain. Available datasets for 3D Action Recognition generally feature controlled lighting and framing while recording subjects from the front. They mostly reflect good recording conditions and therefore fail to account for the highly variable conditions the robot would have to work with in the field, e.g. when providing in-field logistic support for human fruit pickers as in our scenario. Existing work on Intention Recognition mostly labels plans or actions as intentions, but neither of those fully capture the extend of human intent. In this work, we argue for a holistic view on human Intention Recognition and propose a set of recording conditions, gestures and behaviors that better reflect the environment and conditions an agricultural robot might find itself in. We demonstrate the utility of the dataset by means of evaluating two human detection methods: Bounding boxes and skeleton extraction.

Modern smartphones can provide a multitude of services to assist people with visual impairments, and their cameras in particular can be useful for assisting with tasks, such as reading signs or searching for objects in unknown environments. Previous research has looked at ways to solve these problems by processing the camera’s video feed, but very little work has been done in actively guiding the user towards specific points of interest, maximising the effectiveness of the underlying visual algorithms. In this paper, we propose a control algorithm based on a Markov Decision Process that uses a smartphone’s camera to generate real-time instructions to guide a user towards a target object. The solution is part of a more general active vision application for people with visual impairments. An initial implementation of the system on a smartphone was experimentally evaluated with participants with healthy eyesight to determine the performance of the control algorithm. The results show the effectiveness of our solution and its potential application to help people with visual impairments find objects in unknown environments.

This research addresses the challenging problem of visual collision detection in very complex and dynamic real physical scenes, specifically, the vehicle driving scenarios. This research takes inspiration from a large-field looming sensitive neuron, i.e., the lobula giant movement detector (LGMD) in the locust’s visual pathways, which represents high spike frequency to rapid approaching objects. Building upon our previous models, in this paper we propose a novel inhibition mechanism that is capable of adapting to different levels of background complexity. This adaptive mechanism works effectively to mediate the local inhibition strength and tune the temporal latency of local excitation reaching the LGMD neuron. As a result, the proposed model is effective to extract colliding cues from complex dynamic visual scenes. We tested the proposed method using a range of stimuli including simulated movements in grating backgrounds and shifting of a natural panoramic scene, as well as vehicle crash video sequences. The experimental results demonstrate the proposed method is feasible for fast collision perception in real-world situations with potential applications in future autonomous vehicles.

The use of unmanned aerial vehicles (UAVs) in precision agriculture is gaining more and more interest. In this paper, we present a deep learning based method for estimating the crop and weed distribution from images captured by a UAV. The proposed approach runs on an embedded board equipped with a GPU. Quantitative experimental results have been obtained using real images from two different public datasets. The results demonstrate the effectiveness of the proposed approach.

Spinal cord injury can cause cognitive impairments even when no cerebral lesion is appreciable. As patients are forced to explore the environment in a non-canonical position (i.e., seated on a wheelchair), a modified relation with space can explain motor-related cognitive differences compared to non-injured individuals. Peripersonal space is encoded in motor terms, that is, in relation to the representation of action abilities and is strictly related to the affordance of reachability. In turn, affordances, the action possibilities suggested by relevant properties of the environment, are related to the perceiver’s peripersonal space and motor abilities. One might suppose that these motor-related cognitive abilities are compromised when an individual loses the ability to move. We shed light on this issue in 10 patients with paraplegia and 20 matched controls. All have been administered an affordances-related reachability judgement task adapted from Costantini, Ambrosini, Tieri, Sinigaglia, and Committeri (2010, Experimental Brain Research, 207, 95) and neuropsychological tests. Our findings demonstrate that patients and controls show the same level of accuracy in estimating the location of their peripersonal space boundaries, but only controls show the typical overestimation of reaching range. Secondly, patients show a higher variability in their judgements than controls. Importantly, this finding is related to the patients’ ability to perform everyday tasks. Finally, patients are not faster in making their judgements on reachability in peripersonal space, while controls are. Our results suggest that not moving freely or as usual in the environment impact decoding of action-related properties even when the upper limbs are not compromised.

Despite the growing interest in brain–machine interface (BMI)-driven neuroprostheses, the translation of the BMI output into a suitable control signal for the robotic device is often neglected. In this article, we propose a novel control approach based on dynamical systems that was explicitly designed to take into account the nature of the BMI output that actively supports the user in delivering real-valued commands to the device and, at the same time, reduces the false positive rate. We hypothesize that such a control framework would allow users to continuously drive a mobile robot and it would enhance the navigation performance. 13 healthy users evaluated the system during three experimental sessions. Users exploit a 2-class motor imagery BMI to drive the robot to five targets in two experimental conditions: with a discrete control strategy, traditionally exploited in the BMI field, and with the novel continuous control framework developed herein. Experimental results show that the new approach: 1) allows users to continuously drive the mobile robot via BMI; 2) leads to significant improvements in the navigation performance; and 3) promotes a better coupling between user and robot. These results highlight the importance of designing a suitable control framework to improve the performance and the reliability of BMI-driven neurorobotic devices.

We analyze non-linear interactions of 2N -form Maxwell fields in a space-time of dimension D = 4N.Based on the Pasti-Sorokin-Tonin (PST) method, we derive the general consistency condition for the dynamics to respect both manifest SO(2)-duality invariance and manifest Lorentz invariance. For a generic dimension D = 4N, we determine a canonical class of exact solutions of this condition, which represent a generalization of the known non-linear duality invariant Maxwell theories in D = 4. The resulting theories are shown to be equivalent to a corresponding class of canonical theories formulated à la Gaillard-Zumino-Gibbons-Rasheed (GZGR), where duality is a symmetry only of the equations of motion. In dimension D = 8, via a complete solution of the PST consistency condition, we derive new non-canonical manifestly duality invariant quartic interactions. Correspondingly, we construct new non-trivial quartic interactions also in the GZGR approach, and establish their equivalence with the former. In the presence of charged dyonic p-brane sources, we reveal a basic physical inequivalence of the two approaches. The power of our method resides in its universal character, reducing the construction of non-linear duality invariant Maxwell theories to a purely algebraic problem.

Introduction: Patients with late-onset Pompe disease (LOPD) receiving enzyme replacement therapy (ERT) may develop IgG antibodies against alglucosidase alpha (anti-rhGAA) in the first 3 months of treatment. The exact role of these antibodies in modulating efficacy of ERT in this group of patients is still not fully understood. To assess whether anti rh-GAA antibodies interfere with ERT efficacy, we studied a large Italian cohort of LOPD patients. Methods: We analyzed clinical findings and performed serial measurements of IgG anti rh-GAA antibody titers from 64 LOPD patients treated with ERT. The first examination (T0) was completed on average at 17.56 months after starting ERT, while the follow-up (T1) was collected on average at 38.5 months. Differences in T0-T1 delta of the six-minute walking test (6MWT), MRC sum score (MRC), gait, stairs and chair performance (GSGC) and forced vital capacity (FVC) were considered and then related to the antibody titers. Results: Almost 22% of the patients never developed antibodies against GAA, while 78.1% had a positive titer (31.2% patients developed a low titer, 43.8% a medium titer and 3.1% a high titer). No statistical significance was found in relating the T0-T1 delta differences and antibody titers, except for MRC sum score values in a subgroup of patients treated36 months, in which those with a null antibody titer showed a greater clinical improvement than patients with a positive titer. Conclusion: Our results confirm that in a large cohort of LOPD patients, anti rh-GAA antibody generation did not significantly affect either clinical outcome or ERT efficacy. However, in the first 36 months of treatment, a possible interference of low-medium antibody titers with the clinical status could be present. Therefore, a careful and regular evaluation of antibody titers, especially in cases with evidence of clinical decline despite ERT, should be performed.

Human detection and tracking is an essential task for service robots, where the combined use of multiple sensors has potential advantages that are yet to be fully exploited. In this paper, we introduce a framework allowing a robot to learn a new 3D LiDAR-based human classifier from other sensors over time, taking advantage of a multisensor tracking system. The main innovation is the use of different detectors for existing sensors (i.e. RGB-D camera, 2D LiDAR) to train, online, a new 3D LiDAR-based human classifier based on a new ‘trajectory probability’. Our framework uses this probability to check whether new detection belongs to a human trajectory, estimated by different sensors and/or detectors, and to learn a human classifier in a semi-supervised fashion. The framework has been implemented and tested on a real-world dataset collected by a mobile robot. We present experiments illustrating that our system is able to effectively learn from different sensors and from the environment, and that the performance of the 3D LiDAR-based human classification improves with the number of sensors/detectors used.