Costas S. Tzafestas | National Technical University of Athens (original) (raw)

Papers by Costas S. Tzafestas

This paper presents ORB-LINE-SLAM, a real-time hybrid point-line and only-line based visual SLAM ... more This paper presents ORB-LINE-SLAM, a real-time hybrid point-line and only-line based visual SLAM system for stereo cameras which can operate in standard CPUs. This work is based on the ORB-SLAM3 open-source library, which is here expanded and adapted to further cope with line features. The main contribution of this paper concerns the introduction of an adaptation scheme, based on an experimentally tuned adapting factor, which aims to achieve a more efficient fusion of point and line features during the SLAM process. Furthermore, to the best of our knowledge, this work constitutes the first open-source visual SLAM system that has the ability to perform effectively by using exclusively line features. This paper also performs a systematic comparison of different error functions used in the bundle adjustment process, to deduce which ones are the more efficient for the visual SLAM problem. The experimental results obtained indicate that the proposed point-line method significantly improv...

2022 International Conference on Robotics and Automation (ICRA), May 23, 2022

In this paper we introduce a novel technique that aims to control a two-module bio-inspired soft-... more In this paper we introduce a novel technique that aims to control a two-module bio-inspired soft-robotic arm in order to qualitatively reproduce human demonstrations. The main idea behind the proposed methodology is based on the assumption that a complex trajectory can be derived from the composition and asynchronous activation of learned parameterizable simple movements constituting a knowledge base. The present work capitalises on recent research progress in Movement Primitive (MP) theory in order to initially build a library of Probabilistic MPs (ProMPs), and subsequently to compute on the fly their proper combination in the task space resulting in the requested trajectory. At the same time, a model learning method is assigned with the task to approximate the inverse kinematics, while a replanning procedure handles the sequential and/or parallel ProMPs' asynchronous activation. Taking advantage of the mapping at the primitive-level that the ProMP framework provides, the composition is transferred into the actuation space for execution. The proposed control architecture is experimentally evaluated on a real soft-robotic arm, where its capability to simplify the trajectory control task for robots of complex unmodeled dynamics is exhibited.

Fast adaptation to changes in the environment requires both natural and artificial agents to be a... more Fast adaptation to changes in the environment requires both natural and artificial agents to be able to dynamically tune an exploration-exploitation trade-off during learning. This trade-off usually determines a fixed proportion of exploitative choices (i.e. choice of the action that subjectively appears as best at a given moment) relative to exploratory choices (i.e. testing other actions that now appear worst but may turn out promising later). The problem of finding an efficient exploration-exploitation trade-off has been well studied both in the Machine Learning and Computational Neuroscience fields. Rather than using a fixed proportion, non-stationary multi-armed bandit methods in the former have proven that principles such as exploring actions that have not been tested for a long time can lead to performance closer to optimalbounded regret. In parallel, researches in the latter have investigated solutions such as progressively increasing exploitation in response to improvements of performance, transiently increasing exploration in response to drops in average performance, or attributing exploration bonuses specifically to actions associated with high uncertainty in order to gain information when performing these actions. In this work, we first try to bridge some of these different methods from the two research fields by rewriting their decision process with a common formalism. We then show numerical simulations of a hybrid algorithm combining bio-inspired meta-learning, kalman filter and exploration bonuses compared to several state-of-the-art alternatives on a set of non-stationary stochastic multi-armed bandit tasks. While we find that different methods are appropriate in different scenarios, the hybrid algorithm displays a good combination of advantages from different methods and outperforms these methods in the studied scenarios.

IEEE Transactions on Robotics, Apr 1, 2023

HAL (Le Centre pour la Communication Scientifique Directe), Sep 2, 2017

International Journal of Computer Assisted Radiology and Surgery, Apr 24, 2023

Purpose Using robotic technology and communications infrastructure to remotely perform surgery ha... more Purpose Using robotic technology and communications infrastructure to remotely perform surgery has been a persistent goal in medical research in the past three decades. The recent deployment of the Fifth-Generation Wireless Networks has revitalized the research efforts in the telesurgery paradigm. Offering low latency and high bandwidth communication, they are well suited for applications that require real-time data transmission and can allow smoother communication between surgeon and patient, making it possible to remotely perform complex surgeries. In this paper, we investigate the effects of the 5 G network on surgical performance during a telesurgical demonstration where the surgeon and the robot are separated by nearly 300 km. Methods The surgeon performed surgical exercises on a robotic surgery training phantom using a novel telesurgical platform. The master controllers were connected to the local site on a 5 G network, teleoperating the robot remotely in a hospital. A video feed of the remote site was also streamed. The surgeon performed various tasks on the phantom such as cutting, dissection, pick-and-place and ring tower transfer. To assess the usefulness, usability and image quality of the system, the surgeon was subsequently interviewed using three structured questionnaires. Results All tasks were completed successfully. The low latency and high bandwidth of the network resulted into a latency of 18 ms for the motion commands while the video delay was about 350 ms. This enabled the surgeon to operate smoothly with a high-definition video from about 300 km away. The surgeon viewed the system's usability in a neutral to positive way while the video image was rated as of good quality. Conclusion 5 G networks provide significant advancement in the field of telecommunications, offering faster speeds and lower latency than previous generations of wireless technology. They can serve as an enabling technology for telesurgery and further advance its application and adoption.

Fast adaptation to changes in the environment requires agents (animals, robots and simulated arte... more Fast adaptation to changes in the environment requires agents (animals, robots and simulated artefacts) to be able to dynamically tune an exploration-exploitation trade-off during learning. This trade-off usually determines a fixed proportion of exploitative choices (i.e. choice of the action that subjectively appears as best at a given moment) relative to exploratory choices (i.e. testing other actions that now appear worst but may turn out promising later). Rather than using a fixed proportion, nonstationary multi-armed bandit methods in the field of machine learning have proven that principles such as exploring actions that have not been tested for a long time can lead to performance closer to optimal-bounded regret. In parallel, researches in active exploration in the fields of robot learning and computational neuroscience of learning and decision-making have proposed alternative solutions such as transiently increasing exploration in response to drops in average performance, or attributing exploration bonuses specifically to actions associated with high uncertainty in order to gain information when choosing them. In this work, we compare different methods from machine learning, computational neuroscience and robot learning on a set of nonstationary stochastic multi-armed bandit tasks: abrupt shifts; best bandit becomes worst one and vice versa; multiple shifting frequencies. We find that different methods are appropriate in different scenarios. We propose a new hybrid method combining bio-inspired meta-learning, kalman filter and exploration bonuses and show that it outperforms other methods in these scenarios.

arXiv (Cornell University), Dec 1, 2018

In this work, we present a novel framework for on-line human gait stability prediction of the eld... more In this work, we present a novel framework for on-line human gait stability prediction of the elderly users of an intelligent robotic rollator using Long Short Term Memory (LSTM) networks, fusing multimodal RGB-D and Laser Range Finder (LRF) data from non-wearable sensors. A Deep Learning (DL) based approach is used for the upper body pose estimation. The detected pose is used for estimating the body Center of Mass (CoM) using Unscented Kalman Filter (UKF). An Augmented Gait State Estimation framework exploits the LRF data to estimate the legs' positions and the respective gait phase. These estimates are the inputs of an encoder-decoder sequence to sequence model which predicts the gait stability state as Safe or Fall Risk walking. It is validated with data from real patients, by exploring different network architectures, hyperparameter settings and by comparing the proposed method with other baselines. The presented LSTMbased human gait stability predictor is shown to provide robust predictions of the human stability state, and thus has the potential to be integrated into a general user-adaptive control architecture as a fall-risk alarm.

For a context-aware robotic assistant platform that follows patients with moderate mobility impai... more For a context-aware robotic assistant platform that follows patients with moderate mobility impairment and adapts its motion to the patient's needs, the development of an efficient leg tracker and the recognition of pathological gait are very important. In this work, we present the basic concept for the robot control architecture and analyse three essential parts of the Adaptive Context-Aware Robot Control scheme; the detection and tracking of the subject's legs, the gait modelling and classification and the computation of gait parameters for the impairment level assessment. We initially process raw laser data and estimate the legs' position and velocity with a Kalman Filter and then use this information as input for a Hidden Markov Model-based framework that detects specific gait patterns and classifies human gait into normal or pathological. We then compute gait parameters commonly used for medical diagnosis. The recognised gait patterns along with the gait parameters will be used for the impairment level assessment, which will activate certain control assistive actions regarding the pathological state of the patient.

International Conference on Robotics and Automation, 2016

Mobility assistance robots (MARs) provide support to elderly or patients during walking. The desi... more Mobility assistance robots (MARs) provide support to elderly or patients during walking. The design of a safe and intuitive assistance behavior is one of the major challenges in this context. Here we present work on two modes of physical Human-Robot interaction; one where the human in is direct contact with the MAR, e.g. by holding some handles, and the other where the human releases the handles whilst the MAR has to follow him/her from the front, i.e. contactless. For the first mode, we present an integrated approach for the context-specific, on-line adaptation of the assistance level of a rollator-type MAR by gain-scheduling of low-level robot control parameters. A human-inspired decision-making model, the Drift-Diffusion Model, is introduced as the key principle to gain-schedule parameters and with this to adapt the provided robot assistance in order to achieve a human-like assistive behavior. The MAR is designed to provide a) cognitive assistance to help the user follow a desired path as well as b) sensorial assistance to avoid collisions with obstacles while allowing for an intentional approach of them. For the second mode, an intention-based assistive controller for allowing the robot to follow a human while moving in the front is analysed. This task is particularly challenging in indoor environments, as there are situations that are undecidable, namely in junctions. We describe a novel local kinodynamic planner which concurrently detects discrete routes and continuous motion paths. An intention recognition algorithm is also detailed, along with tests in a T-Junction.

Using robots as therapeutic or educational tools for children with autism requires robots to be a... more Using robots as therapeutic or educational tools for children with autism requires robots to be able to adapt their behavior specifically for each child with whom they interact. In particular, some children may like to be looked into the eyes by the robot while some may not. Some may like a robot with an extroverted behavior while others may prefer a more introverted behavior. Here we present an algorithm to adapt the robot's expressivity parameters of action (mutual gaze duration, hand movement expressivity) in an online manner during the interaction. The reward signal used for learning is based on an estimation of the child's mutual engagement with the robot, measured through non-verbal cues such as the child's gaze and distance from the robot. We first present a pilot joint attention task where children with autism interact with a robot whose level of expressivity is predetermined to progressively increase, and show results suggesting the need for online adaptation of expressivity. We then present the proposed learning algorithm and some promising simulations in the same task. Altogether, these results suggest a way to enable robot learning based on non-verbal cues and to cope with the high degree of nonstationarities that can occur during interaction with children.

Elderly people have particular needs in performing bathing activities, since these tasks require ... more Elderly people have particular needs in performing bathing activities, since these tasks require body flexibility. Our aim is to build an assistive robotic bath system, in order to increase the independence and safety of this procedure. Towards this end, the expertise of professional carers for bathing sequences and appropriate motions has to be adopted, in order to achieve natural, physical human-robot interaction. In this paper, a real-time end-effector motion planning method for an assistive bath robot, using on-line Point-Cloud information, is proposed. The visual feedback obtained from Kinect depth sensor is employed to adapt suitable washing paths to the user's body part motion and deformable surface. We make use of a navigation function-based controller, with guarantied globally uniformly asymptotic stability, and bijective transformations for the adaptation of the paths. Experiments were conducted with a rigid rectangular object for validation purposes, while a female subject took part to the experiment in order to evaluate and demonstrate the basic concepts of the proposed methodology.

Human Factors in Robots, Drones and Unmanned Systems

This paper presents assessment findings of the “i-Walk” robotic rehabilitation assistant. i-Walk ... more This paper presents assessment findings of the “i-Walk” robotic rehabilitation assistant. i-Walk provides support to target groups of people with cognitive and/or mobility deficits via a pioneer robotic rollator that utilizes innovation in multimodal robot perception, user-adaptive robot autonomy and natural human-robot interaction. The i-Walk rollator was thoroughly evaluated in terms of its usability and acceptance from its intended end users (patients and therapists) in a rehabilitation centre. i-Walk was tested (i) as a whole, and in terms of (ii) its navigation and human-robot interaction functionalities, (iii) the provided walking support, and (iv) the rehabilitation exercises it offers. In total, twenty-two patients and twelve therapists evaluated the device under real conditions. The paper presents the findings from the evaluation testing of the i-Walk platform. A systematic methodology and protocol were used to test the intelligent robotic rehabilitation assistant in three ...

This paper presents a contribution to the active field of robotics research to support the develo... more This paper presents a contribution to the active field of robotics research to support the development of social skills and capabilities in children with Autism Spectrum Disorders as well as Typically Developing children. We present preliminary results of a novel experiment where classical roles are reversed: children are here the teachers giving positive or negative reinforcement to the Kaspar robot to make it learn arbitrary associations between toys and locations where to tidy them. The goal is to help children change perspective, and understand that sometimes a learning agent needs several repetitions before correctly learning something. We developed a reinforcement learning algorithm enabling Kaspar to verbally convey its uncertainty along learning, so as to better inform the interacting child of the reasons behind successes and failures made by the robot. Overall, 30 children aged between 7 and 8 (19 girls, 11 boys) performed 16 sessions of the experiment in groups, and manage...

International Journal of Social Robotics, 2019

This paper presents a contribution aiming at testing novel child–robot teaching schemes that coul... more This paper presents a contribution aiming at testing novel child–robot teaching schemes that could be used in future studies to support the development of social and collaborative skills of children with autism spectrum disorders (ASD). We present a novel experiment where the classical roles are reversed: in this scenario the children are the teachers providing positive or negative reinforcement to the Kaspar robot in order for it to learn arbitrary associations between different toy names and the locations where they are positioned. The objective is to stimulate interaction and collaboration between children while teaching the robot, and also provide them tangible examples to understand that sometimes learning requires several repetitions. To facilitate this game, we developed a reinforcement learning algorithm enabling Kaspar to verbally convey its level of uncertainty during the learning process, so as to better inform the children about the reasons behind its successes and failu...

Lecture Notes in Computer Science, 2016

Acquisition and annotation of a multimodal-multisensory data set of human-passive rollator-carer ... more Acquisition and annotation of a multimodal-multisensory data set of human-passive rollator-carer interactions have enabled the analysis of related human behavioural patterns and the definition of the MOBOT human-robot communication model. The MOBOT project has envisioned the development of cognitive robotic assistant prototypes that act proactively, adaptively and interactively with respect to elderly humans with slight walking and cognitive difficulties. To meet the project’s goals, a multimodal action recognition system is being developed to monitor, analyse and predict user actions with a high level of accuracy and detail. In the same framework, the analysis of human behaviour data that have become available through the project’s multimodal-multisensory corpus, have led to the modelling of Human-Robot Communication in order to achieve an effective, natural interaction between users and the assistive robotic platform. Here, we discuss how the project’s communication model has been integrated in the robotic platform in order to support a natural multimodal human-robot interaction.