Search

Advances in the automotive industry are changing the relationship between cars and drivers. Advanced driver assistant systems, such as navigation systems, advanced cruise control, collision avoidance systems, and other safety systems, are now common and assist the driver in controlling the car. Smart key fobs have made getting in and starting the car almost effortless: the fob can be left in the pocket and the doors will unlock when a driver/user approaches the car and simply touches the door handle. This is a level of personalization and convenience that is almost standard today. The research presented here brings a new perspective on personalization and driver assistance systems. An online survey was conducted, which aimed to gather public opinion on the usefulness of endowing future (semi-)autonomous cars with social and cognitive behavior, such as the ability to learn drivers' routines and preferences in order to make decisions and perform actions in preparation for the next trip and to manage comfort within the cockpit without being commanded to do so. After filtering, the study included 657 respondents from 93 nations. The results demonstrate a favorable opinion towards such human-centered cognitive vehicles and could be helpful for designers in the automotive industry and other related stakeholders in the development of future cognitive vehicles.

Many of our sequential activities require that behaviors must be both precisely timed and put in the proper order. This article presents a neurocomputational model based on the theoretical framework of dynamic neural fields that supports the rapid learning and flexible adaptation of coupled order-timing representations of sequential events. A key assumption is that elapsed time is encoded in the monotonic buildup of self-stabilized neural population activity representing event memory. A stable activation gradient over subpopulations carries the information of an entire sequence. With robotics applications in mind, we test the model in simulations of learning by observation paradigm, in which the cognitive agent first memorizes the order and relative timing of observed events and, subsequently, recalls the information from memory taking potential speed constraints into account. Model robustness is tested by systematically varying sequence complexity along the temporal and the ordinal dimensions. Furthermore, an adaptation rule is proposed that allows the agent to adjust in a single trial a learned timing pattern to a changing temporal context. The simulation results are discussed with respect to our goal to endow autonomous robots with the capacity to efficiently learn complex sequences with time constraints, supporting more natural human-robot interactions., This work was supported in part by FCT (Portuguese Foundation for Science and Technology) through the Ph.D. Fellowship under Grant PD/BD/128183/2016; in part by the European Structural and Investment Funds in the FEDER Component, through the Operational Competitiveness and Internationalization Programme (COMPETE 2020) and National Funds, through the FCT under Project PTDC/MAT-APL/31393/2017 (NEUROFIELD) and Project POCI-01-0247FEDER-039334; and in part by Research and Development Units Project Scope under Project UIDB/00319/2020 and Project UIDB/00013/2020.

Diagnosis of Fabry disease (FD) remains a challenge mostly due to its rare occurrence and phenotipical variability, with considerable delay between onset and clinical diagnosis. It is then of extreme importance to explore biomarkers capable of assisting the earlier diagnosis of FD. There is growing evidence supporting the use of gait assessment in the diagnosis and management of several neurological diseases. In fact, gait abnormalities have previously been observed in FD, justifying further investigation. The aim of this study is to evaluate the effectiveness of different machine learning strategies when distinguishing patients with FD from healthy controls based on normalized gait features. Gait features of an individual are affected by physical characteristics including age, height, weight, and gender, as well as walking speed or stride length. Therefore, in order to reduce bias due to inter-subject variations a multiple regression (MR) normalization approach for gait data was performed. Four different machine learning strategies - Support Vector Machines (SVM), Random Forest (RF), Multiple Layer Perceptrons (MLPs), and Deep Belief Networks (DBNs) - were employed on raw and normalized gait data. Wearable sensors positioned on both feet were used to acquire the gait data from 36 patients with FD and 34 healthy subjects. Gait normalization using MR revealed significant differences in percentage of stance phase spent in foot flat and pushing ( $p ), with FD presenting lower percentages in foot flat and higher in pushing. No significant differences were observed before gait normalization. Support Vector Machine was the superior classifier achieving an FD classification accuracy of 78.21% after gait normalization, compared to 71.96% using raw gait data. Gait normalization improved the performance of all classifiers. To the best of our knowledge, this is the first study on gait classification that includes patients with FD, and our results support the use of gait assessment on the clinical assessment of FD.

Differential diagnosis between Idiopathic Parkin-son's disease (IPD) and Vascular Parkinsonism (VaP) is a difficult task, especially early in the disease. There is growing evidence to support the use of gait assessment in diagnosis and management of movement disorder diseases. The aim of this study is to evaluate the effectiveness of some machine learning strategies in distinguishing IPD and VaP gait. Wearable sensors positioned on both feet were used to acquire the gait data from 15 IPD, 15 VaP, and 15 healthy subjects. A comparative classification analysis was performed by applying two supervised machine learning algorithms: Multiple Layer Perceptrons (MLPs) and Deep Belief Networks (DBNs). The decisional space was composed of the gait variables, with or without neuropsychological evaluation (Montreal cognitive assessment (MoCA) score), top-ranked in an error incremental analysis. In the classification task of characterizing parkinsonian gait by distinguishing between patients (IPD+VaP) and healthy control, from the all strides classification of the gait performed by the person, high accuracy (93% with or without MoCA) was obtained for both algorithms. In the classification task of the two groups of patients (VaP vs. IPD), DBN classifier achieved higher performance (73% with MoCA). To the best of our knowledge, this is the first study on gait classification that includes a VaP group. DBN classifiers are not frequently applied in literature to similar studies, but the results here obtained demonstrate that the use of DBN classifiers based on gait analysis is promising to be a good support to the neurologist in distinguishing VaP and IPD.

If we want robots to engage effectively with humans in service applications or in collaborative work scenarios they have be endowed with the capacity to perceive the passage of time and control the timing of their actions. Here we report result of a robotics experiment in which we test a computational model of action timing based on processing principles of neurodynamics. A key assumption is that elapsed time is encoded in the consistent buildup of persistent population activity representing the memory of sensory or motor events. The stored information can be recalled using a ramp-to-threshold dynamics to guide actions in time. For the experiment we adopt an assembly paradigm from our previous work on natural human-robot interactions. The robot first watches a human executing a sequence of assembly steps. Subsequently, it has to execute the steps from memory in the correct order and in synchrony with an external timing signal. We show that the robot is able to efficiently adapt its motor timing and to store this information in memory using the temporal mismatch between the neural processing of the sensory feedback about executed actions and the external cue., FCT - Fundació Catalana de Trasplantament(PD/BD/128183/2016)This research was supported by the Marie Curie Network for Initial Training NETT, FCT through the PhD fellowship PD/BD/128183/2016, the FCT-Research Center CMAT (PEstOE/MAT/UI0013/2014), and FCT - Algoritmi research Centre (COMPETE: POCI-01-0145-FEDER-007043 and FCT - Fundação para a Ciência e Tecnologia within the Project ˆ Scope: UID/CEC/00319/2013)

Many of our everyday tasks require the control of the serial order and the timing of component actions. Using the dynamic neural field (DNF) framework, we address the learning of representations that support the performance of precisely time action sequences. In continuation of previous modeling work and robotics implementations, we ask specifically the question how feedback about executed actions might be used by the learning system to fine tune a joint memory representation of the ordinal and the temporal structure which has been initially acquired by observation. The perceptual memory is represented by a self-stabilized, multi-bump activity pattern of neurons encoding instances of a sensory event (e.g., color, position or pitch) which guides sequence learning. The strength of the population representation of each event is a function of elapsed time since sequence onset. We propose and test in simulations a simple learning rule that detects a mismatch between the expected and realized timing of events and adapts the activation strengths in order to compensate for the movement time needed to achieve the desired effect. The simulation results show that the effector-specific memory representation can be robustly recalled. We discuss the impact of the fast, activation-based learning that the DNF framework provides for robotics applications.

We tested in a robotics experiment a dynamic neural field model for learning a precisely timed musical sequence. Based on neuro-plausible processing mechanisms, the model implements the idea that order and relative timing of events are stored in an integrated representation whereas the onset of sequence production is controlled by a separate process. Dynamic neural fields provide a rigorous theoretical framework to analyze and implement the necessary neural computations that bridge gaps between sensation and action in order to mediate working memory, action planing, and decision making. The robot first memorizes a short musical sequence performed by a human teacher by watching color coded keys on a screen, and then tries to execute the piece of music on a keyboard from memory without any external cues. The experimental results show that the robot is able to correct in very few demonstration-execution cycles initial sequencing and timing errors., The work received financial support from FCT - Fundação para a Ciência e Tecnologia within the Project Scope: PEst- OE/EEI/UI0319/2014, the Research Centers for Mathematics and Algoritmi through the FCT Pluriannual Funding Program, PhD and Post-doctoral Grants (SFRH/BD/41179/2007, SFRH/BD/48529/2008 and SFRH/BPD/71874/2010, financed by POPH-QREN-Type 4.1-Advanced Training, co-funded by the European Social Fund and national funds from MEC), and Project NETT: Neural Engineering Transformative Technologies, EU-FP7 ITN (nr.289146).

Humans are experts in cooperating in a smooth and proactive manner. Action and intention understanding are critical components of efficient joint action. In the context of the EU Integrated Project JAST [16] we have developed an anthropomorphic robot endowed with these cognitive capacities. This project and respective robot (ARoS) is the focus of the video. More specifically, the results illustrate crucial cognitive capacities for efficient and successful human-robot collaboration such as goal inference, error detection and anticipatory action selection. Results were considered one of the ICT "success stories", JAST: Joint-Action Science and Technology” (Ref. IST-2-003747-IP), FCT FCOMP-01-0124-FEDER-022674”.