The development of autonomous robotic systems offers significant potential for performing complex tasks with precision and consistency. Recent advances in Artificial Intelligence (AI) have enabled more capable intelligent automation systems, addressing increasingly complex challenges. However, this progress raises questions about human roles in such systems. Human-Centered AI (HCAI) aims to balance human control and automation, ensuring performance enhancement while maintaining creativity, mastery, and responsibility. For real-world applications, autonomous robots must balance task performance with reliability, safety, and trustworthiness. Integrating HCAI principles enhances human-robot collaboration and ensures responsible operation.This paper presents a bibliometric analysis of intelligent autonomous robotic systems, utilizing SciMAT and VOSViewer to examine data from the Scopus database. The findings highlight academic trends, emerging topics, and AI’s role in self-adaptive robotic behaviour, with an emphasis on HCAI architecture. These insights are then projected onto the IBM MAPE-K architecture, with the goal of identifying how these research results map into actual robotic autonomous systems development efforts for real-world scenarios.

Autonomous decision-making has always been one of the primary goals to pursue as concerns mobile robots. Researchers of this field have recently turned their attention to Deep Reinforcement Learning (DRL). This paper presents a Double Deep Q Network architecture for managing the high level decisions of a mobile robot involved in a site servicing task. We imagined a scenario where an autonomous service robot must react to alarms due to failures in its area of interest; the robot must have onboard the necessary servicing tool by resorting to a tool change station if needed, reach the area of the failure and fix it, while at the same time handling its battery status. One of the key properties, yet rarely examined, when it comes to robots' long-term independence is the energy-awareness, namely the ability of autonomously managing the charge state as a function of current and future needs. The proposed Deep Q Network training reward scheme is defined specifically to obtain an energy-aware high-level controller, by penalizing both extremely low levels of battery charge as well as unnecessary recharges. The model is numerically simulated on a graph scenario constituted of several failure and charging nodes. Results show that the trained agent always succeeds in reaching the destination without ever incurring in a complete discharge, as it promptly performs temporary stops at charging locations whenever needed.

This paper revises the SISO output feedback$\mathcal{L}_{1}$adaptive control architecture and the$\mathcal{L}_{1}$adaptive augmentation of a baseline controller. An improved stability condition for the baseline$\mathcal{L}_{1}$augmentation architecture is derived. The proposed improved way of performing the stability analysis can be less conservative, and this result may help certify the baseline augmentation for a broader range of uncertainties and disturbances.

The paper presents an overview and the main results of the Erasmus+ project titled Interactive Course for Control Theory – ICCT. The project was started by four European Universities when they identified by a survey that students find Control Theory more challenging to understand compared to other subjects. The project’s main goal was to develop a platform, which would offer examples from the field of Control Theory in an interactive way – ICCT. Frequently asked questions (FAQ) and test exam questions were added as well. To support the course, an accompanying textbook was prepared. At this point, the course and the accompanying textbook are available in English and the languages of the partner universities; since both of them are publicly available, we assume they will be offered in other languages in future. The benefits of implementing our interactive course are already visible in better survey results and improved grades in Control Theory-related subjects.

This paper describes the process of design, analysis, and scoring with both subjective and objective criteria, of a set of possible haptic inceptors configurations to be used as primary flight control devices for a class of future tilt rotor vehicles. Several scoring criteria were adopted like ergonomics, cognitive compatibility, easiness of recovery in fault situations, and expected workload among several others. Each criterion was weighted to obtain a scoring matrix for a total of 18 different configurations, some "classic", some novel, leading to an overall ranking. The matrix was used to select the best candidate configurations to be tested using a human-in-the-loop simulator in the context of the EU Funded CleanSky2 SMARTeR Project. After a final review phase, the best solution will be adopted to actually fly the Leonardo NextGen Civil Tilt Rotor Technological Demonstrator in the context of the EU Funded CleanSky2 SAIS Project.

This paper presents the experimental evaluation of a novel haptic shared control system. The proposed system was designed to help human operators to perform a maneuver that is not known in advance. A Pilot Intent Estimator (PIE) was implemented to estimate the unknown trajectory that a human pilot intends to follow. Then, a haptic feedback was designed to mimic the behavior of a skilled pilot that tracks the estimated trajectory. A human-in-the-loop experiment was performed to evaluate the developed shared control system in a 2 Degrees-of- Freedom (DoF) control task. Results showed the effectiveness of the PIE to estimate the correct pilot intended path. Furthermore, the developed haptic system helped participants to achieve better performance to follow the estimated trajectory, compared to manual control.

This paper investigates the use of a variable haptic feedback for training a disturbance rejection task. The haptic feedback was designed as a Force-Stiffness feedback. Throughout the training, Force and Stiffness feedback are decreased to progressively give more control authority to the human operator. The training method was tested in a human-in-the-loop experiment. In the experiment, participants were split into three groups: variable haptic aid (VHA), constant haptic aid (CHA) and no haptic aid (NoHA). The VHA and CHA groups performed a first training phase with variable and constant haptic feedback respectively, followed by an evaluation phase without external aids. The NoHA group performed the entire experiment without external aids. Results showed that in the training phase both VHA and CHA groups performed better than NoHA group. In the evaluation phase though, only the VHA group obtained better performances than the NoHA group. Specifically, participants were able to quickly recover similar performances to those obtained at the end of the training phase. Thus, the variable haptic training proved to be more effective than the constant haptic training and manual control at helping participants learn the task.

Haptic guidance has previously been employed to improve human performance in control tasks. This paper presents an experiment to evaluate whether haptic feedback can be used to help humans learn a compensatory tracking task. In the experiment, participants were divided into two groups: the haptic group and the no-aid group. The haptic group performed a first training phase with haptic feedback and a second evaluation phase without haptic feedback. The no-aid group performed the whole experiment without haptic feedback. Results indicated that haptic group achieved better performance than the no-aid group during the training phase. Furthermore, performance of haptic group did not worsen in the evaluation phase when the haptic feedback was turned off. Moreover, the no-aid group needed more experimental trials to achieve similar performance to the haptic group. These findings indicate that haptic feedback helped participants learn the task quicker.

Effectiveness of haptic guidance systems depends on how humans adapt their neuromuscular response to the force feedback. A quantitative insight into adaptation of neuromuscular response can be obtained by identifying neuromuscular dynamics. Since humans are likely to vary their neuromuscular response during realistic control scenarios, there is a need for methods that can identify time-varying neuromuscular dynamics. In this work an identification method is developed which estimates the impulse response of time-varying neuromuscular system by using a Recursive Least Squares (RLS) method. The proposed method extends the commonly used RLS-based method by employing the pseudo inverse operator instead of the inverse operator. This results in improved robustness to external noise. The method was validated in a human in-the-loop experiment. The neuromuscular estimates given by the proposed method were more accurate than those obtained with the commonly used RLS-based method.

In this paper, identification methods are proposed to estimate the neuromuscular and visual responses of a multiloop pilot model. A conventional and widely used technique for simultaneous identification of the neuromuscular and visual systems makes use of cross-spectral density estimates. This paper shows that this technique requires a specific noninterference hypothesis, often implicitly assumed, that may be difficult to meet during actual experimental designs. A mathematical justification of the necessity of the noninterference hypothesis is given. Furthermore, two methods are proposed that do not have the same limitations. The first method is based on autoregressive models with exogenous inputs, whereas the second one combines cross-spectral estimators with interpolation in the frequency domain. The two identification methods are validated by offline simulations and contrasted to the classic method. The results reveal that the classic method fails when the noninterference hypothesis is not fulfilled; on the contrary, the two proposed techniques give reliable estimates. Finally, the three identification methods are applied to experimental data from a closed-loop control task with pilots. The two proposed techniques give comparable estimates, different from those obtained by the classic method. The differences match those found with the simulations. Thus, the two identification methods provide a good alternative to the classic method and make it possible to simultaneously estimate human's neuromuscular and visual responses in cases where the classic method fails.

A human-centered design of haptic aids aims at tuning the force feedback based on the effect it has on human behavior. For this goal, a better understanding of the influence of haptic aids on the pilot neuromuscular response becomes crucial. In realistic scenarios, the neuromuscular response can continuously vary depending on many factors, such as environmental factors or pilot fatigue. This paper presents a method that online estimates time-varying neuromuscular dynamics during force-related tasks. This method is based on a Recursive Least Squares (RLS) algorithm and assumes that the neuromuscular response can be approximated by a Finite Impulse Response filter. The reliability and the robustness of the method were investigated by performing a set of Monte-Carlo simulations with increasing level or remnant noise. Even with high level of remnant noise, the RLS algorithm provided accurate estimates when the neuromuscular dynamics were constant or changed slowly. With instantaneous changes, the RLS algorithm needed almost 8s to converge to a reliable estimate. These results seem to indicate that RLS algorithm is a valid tool for estimating online time-varying admittance.

Haptic technology has become a viable way to support operators in vehicular control. This paper investigates two different design philosophies for continuous haptic feedback to support drivers with curve negotiation. The first system, 'direct haptic assistance', is designed to yield best results when the driver gives way to the guidance forces on the steering wheel. The second, 'indirect haptic assistance', is designed to yield best results when the driver counter-acts the forces. The two designs were compared in a driving simulator experiment in which 27 subjects participated. Results show that both systems are helpful in case of low visibility, where the driver lacks sufficient preview of the curves. With normal visibility no improvements on performance were found, for either system. Further experiments are required to investigate the difference between the two approaches.

This article proposes a novel solution to the Pose Estimation problem for Ego-Motion from stereo camera images. The approach uses a nonlinear function, derived from the concept of Gibbs' Entropy, which is robust by nature to the presence of noise and outliers in the visual features. The SIFT algorithm is used to collect and match the features from stereo images. The 3-vectors quaternion parameterization is used to parameterize the rotation matrix, in order to avoid the unit norm constraint in the minimization computation. Simulations and experimental results are presented and compared with the results obtained via the classical Iterative Closest Point approach.

This paper considers the problem of realizing a 6-DOF closed-loop motion simulator by exploiting an anthropomorphic serial manipulator as motion platform. Contrary to standard Stewart platforms, an industrial anthropomorphic manipulator offers a considerably larger motion envelope and higher dexterity that let envisage it as a viable and superior alternative. Our work is divided in two papers. In this Part I, we discuss the main challenges in adopting a serial manipulator as motion platform, and thoroughly analyze one key issue: the design of a suitable inverse kinematics scheme for online motion reproduction. Experimental results are proposed to analyze the effectiveness of our approach. Part II [1] will address the design of a motion cueing algorithm tailored to the robot kinematics, and will provide an experimental evaluation on the chosen scenario: closed-loop simulation of a Formula 1 racing car.

This paper, divided in two Parts, considers the problem of realizing a 6-DOF closed-loop motion simulator by exploiting an anthropomorphic serial manipulator as motion platform. After having proposed a suitable inverse kinematics scheme in Part I [1], we address here the other key issue, i.e., devising a motion cueing algorithm tailored to the specific robot motion envelope. An extension of the well-known classical washout filter designed in cylindrical coordinates will provide an effective solution to this problem. The paper will then present a thorough experimental evaluation of the overall architecture (inverse kinematics + motion cueing) on the chosen scenario: closed-loop simulation of a Formula 1 racing car. This will prove the feasibility of our approach in fully exploiting the robot motion capabilities as a motion simulator.

The article describes a novel approach for deployment of a swarm of heterogeneous autonomous vehicles. Each vehicle is treated as the agent of a network, which cooperates in order to cover a given area according to its own capabilities. A general framework is introduced, which aims at providing tools for solving a large class of coordination problems. The capabilities of each agent are modeled with Descriptor Functions; the sum of these functions constitutes the swarm's descriptor. The goal of the swarm is to match a desired descriptor, by minimizing an appropriate cost functional. A control law is proposed, which is capable of driving the agents towards the achievement of the goal. The existence of local minima and of a global minimum is discussed. Theoretical results on the existence of the global minimum are given. The Area Coverage problem is selected as a preliminary test for the algorithm. Simulation results show the effectiveness of the proposed approach.

The paper presents an experimental evaluation of two different Haptic aiding concepts: Direct and Indirect Haptic Aiding. Two Haptic systems were designed and tested using an experimental setup. The problem of wind gust rejection in Remotely Piloted Vehicles is used as test bench. Test results show the effectiveness of both methods but a better performance of the IHA-based system for pilots without any previous training about the experiment. DHA-based system provided instead better results after some pilot training on the experiment. Pilots reported better sensation of the wind gusts with IHA-based feedback.