previous day
next day
all days

View: session overviewtalk overview

11:30-12:00 Session 5: Plenary session
Location: Virtual Room A
Human–Robot Collaboration in Industrial Production: Recent Trends and Research at SZTAKI

ABSTRACT. The inclusion of human workers in automated production processes is nowadays receiving growing support by underlying technological and scientific development. Simultaneously, industrial production faces a growing need to reclaim the versatility and intuitive problem solving of human workforce previously lost to industrial automation. While collaborative operation of robots and humans is still subject to intense research with many open issues, the domain has considerably matured and diversified in the past decade. Especially in industrial production, human–robot collaboration is now more and more recognized as a solution portfolio with an entire spectrum from robot-only to human-only cases, known or expected limitations, and ways of integration into a larger-scale production hierarchy. The presentation gives an overview of the current state and key trends in human–robot collaboration in industrial production, and highlights contributions of related research and development at the Research Laboratory of Engineering and Management Intelligence (EMI) of the Institute for Computer Science and Control (SZTAKI).

12:00-14:00 Session 6: Navigation and motion planning
Location: Virtual Room A
3D Head Pointer - Proposal of the manipulation method that enables the spatial localization for a wearable robot arm by head bobbing -

ABSTRACT. This paper proposes a novel interface “3D Head Pointer” to operate spatial localization of a wearable robot arm. The proposed system is expected to support daily tasks (e.g. opening an umbrella, stirring a pot) concurrently with a wearable robot arm. In previous research, it was difficult to operate the robotic arm independently with operator’s natural limbs behavior without interrupting them. The proposed method manipulates spatial localization in a polar coordinate system by combining the orientation of the face and head bobbing, a forward and backward movement of the head. In VR space, we examined the operation magnification of head bobbing and evaluated the prototype system by measuring the instructional accuracy and time. Experiments were conducted on six subjects, the results were confirmed that a high accuracy of 8㎜in 4 seconds of work time and the independence of the operator’s limbs during instruction, it was suggested the effectiveness of the operation method.

Generating instant trajectory of an indoor UAV with respect to its dynamics

ABSTRACT. The application of autonomous indoor UAVs indicates the need for a deeper examination of the trajectory generation. This is due to the fact that the same UAV dynamics are used in more confined space, greatly decreasing the space for maneuvering. The result of path planning methods usually represents a finite set of points with a certain spacing between them. To optimize the system, an effort is made to maximize these distances, while maintaining the shape of the path. To obtain a trajectory suitable for UAV control, it is necessary to define the path between these points analytically. The global trajectory (the whole flight or a significant part of it) can be generated by interpolation or regression of known points. The generation of instant trajectory takes into account the output of reactive navigation and other control systems. The outputs of these systems might provide position targets with large dynamic and directional range over short distances. In our task, we have chosen an approach to generate an instant trajectory via Hermite's spline. With this approach we can generate an instant trajectory from the current state. During generation the transitions between points are continuous even without knowing the previous states, using only the prescribed dynamics at a given point. Therefore, we know the desired position and velocity of the UAV at all times, preventing sharp movements done by a sudden jump of desired position. The result of the task in this paper is the testing and verification of the Hermite’s spline on a real autonomous UAV, which is currently being developed in collaboration with the Institute of Robotics and Cybernetics at the Slovak University of Technology in Bratislava.

Simultaneous localization and mapping on a mobile robot platform
PRESENTER: Martin Lučan

ABSTRACT. Simultaneous localization and mapping (SLAM) is a very current problem in robotics, and the development of new SLAM algorithms is carried out by leading research institutions. The text presents the implementation of selected SLAM algorithms on a custom made hardware platform built on the chassis of the 1:10 remote control Ackerman vehicle model. A transformation of a hobby vehicle platform to an autonomous robot requires hardware and software intervention the paper further describes a sensor setup, hardware configuration, and briefly apprises of the software architecture. Finally, the characteristics and performance of GMapping with wheel odometry and visual odometry, Hector SLAM, and CRSM SLAM algorithms are evaluated in a series of practical experiments.

Comparison of Approaches for UAV Dynamics Consideration in Sampling Based Path Planning Methods
PRESENTER: Jaromir Stanko

ABSTRACT. Deployment of autonomous UAVs in indoor environment puts higher requirements for path planning algorithms due to more complex nature of the environment (narrow crossings, absence of GNSS, smaller manoeuvring space, etc. ...). Due to manoeuvring of the UAV between obstacles it is required to consider its dynamics in the path planning process. Basic versions of sampling-based path planning methods such as RRT (rapidly exploring random trees) or PRM (probabilistic roadmaps) don’t consider dynamics of the object we are planning a path for. In the past, several approaches for path planning with dynamic constraints were introduced. The first approach, randomized kino-dynamic planning, was presented for the first time in [1]. Main goal of this approach is to incrementally search state space as configuration space and find control input that results in collision free trajectory. The second approach, presented in [2], utilizes differential flatness of UAV’s (quadrotor) equations of motion. Thus, it is possible to precisely follow the path previously generated by the path planning algorithm. Main task of this paper is to implement and compare mentioned approaches for path planning with kino-dynamic constraints and decide which one of them is more suitable for the path planning module of autonomous UAV that is currently being developed at the Institute of Robotics and Cybernetics at the Slovak University of Technology in Bratislava. Because of the limited computational power of UAV’s onboard computer, path planning time is one of the parameters we consider. Besides that, path transition time, length and smoothness are considered as well.

[1] LaValle, S. M., & Kuffner, J. J. (2001). Randomized Kinodynamic Planning. The International Journal of Robotics Research, 20(5), 378–400. [2] Richter C., Bry A., Roy N. (2016) Polynomial Trajectory Planning for Aggressive Quadrotor Flight in Dense Indoor Environments. In: Inaba M., Corke P. (eds) Robotics Research. Springer Tracts in Advanced Robotics, vol 114. Springer, Cham

From Coverage Path Planning to parking lot exploration

ABSTRACT. Nowadays, it is getting really difficult to find an adequate free parking space, as there are more and more vehicles travelling on the roads. It is especially a difficult task to find a free parking place in huge parking garages and shopping malls, because the area of the parking zone is really big, and there are a lot of cars searching for a free space. In some parking lots, a parking system is installed. These parking systems contain sensors at every parking space in order to perceive and indicate the occupancy of the given parking space. Using the signals of these systems, the vehicles can be navigated directly to the given parking space, for example using an IoT system. However, most of the parking lots are not equipped with these parking systems, so the driver must circumnavigate them, in order to be able to detect the free parking spaces. For an autonomous vehicle, a path should be planned, which drives along all the possible free parking spaces, so that the sensors of the vehicle could detect them. The parking lot exploration is very similar to the coverage path planning (CPP) problems. This paper presents CPP algorithm based parking lot exploration methods, in which personal preferences can be taken into account, too.

Utilization of collaborative mapping in parking systems

ABSTRACT. Nowadays advanced driver-assistance (ADAS) systems are of general use, as they help vehicle drivers while driving or during parking. These systems can perform the parking tasks when the goal configuration is available for the vehicle. However, the urban environment makes the parking lot search an inevitable and unpleasant task, as for ADAS systems a goal configuration is required in order to perform the parking maneuver. The literature provided solutions for the parking lot exploration mainly focus on the use of preinstalled sensors in parking lots. The presented paper proposes an autonomous parking system using alternately bird's eye view maps and community-built maps such as OpenStreetMap. The alternate use of these maps provides the parking lot exploration subsystem for the autonomous parking system. When an additional environmental sensor is installed on the vehicle, making the detection of parking space candidates possible, the use of preinstalled sensors in parking lots becomes unnecessary.

14:30-15:00 Session 7: Plenary session
Location: Virtual Room A
Unified robot control and navigation in industrial mobile robotics

ABSTRACT. In the industrial mobile robotics sector, the growing need for enhanced modularity and flexibility created the demand for a unified industrial navigation interface through which different AGVs with different characteristics can speak in a uniform manner with a central controller while assuring robustness and extensibility. Due to the high Plug & Play capability assured by such an interface the time needed for the deployment of AGVs to industrial processes reduces dramatically. As an industrial system may include several(1000+) AGVs, optimizing the communication between the AGVs and central control are also key elements of the interface design. The main focus of the presentation is the lately standardized VDA5050 interface serving as a new base for integrating mobile robot and fleet management systems to production processes.  

15:00-16:40 Session 8A: Control and identification
Location: Virtual Room A
Comparison of estimation methods for identification of a robotic flight simulator

ABSTRACT. The use of serial robots for flight simulators implies in several challenges. Among them is the development of the washout filter, which consists on the algorithm that converts the movement of airplane to the limited robot workspace. One strategy to tune the washout filter parameters is based on the use of model-based optimization algorithms. For this purpose, a good model of the dynamic response of motion platform is required. It must receive as inputs of position and orientation in Cartesian space and provide as outputs accelerations and angular velocities in the pilot’s head. In this work, we present and compare different model identification approaches applied to the SIVOR flight simulator model. SIVOR is a robotic simulator that uses a KUKA KR 1000 robot on a KL 1500 linear unit as a motion platform. It is the result of joint work between the Aeronautics Institute of Technology and EMBRAER, the Brazilian aircraft manufacturer. This work complements a previous work dedicated to establishes the causality between input and output signals based on an analysis of coherence and magnitude and, based on this information, propose a model structure. In this work we explore black-box identification techniques aiming to obtain better fitness results and, consequently, a more representative model. The identification is performed using data from two experiments conducted in SIVOR flight simulator. In the first one, only one robot channel was excited during each test-run. In the second one, multiple robot channels were commanded at once. The model identification uses data from the first experiment while the model validation is based on data from the second experiment. The paper brings a comparison between ARX, Box-Jenkins correlation modes and transfer function models aiming to determine the one that results in the best model representativeness.

Eye-to-Hand Architecture for PCB Automated Dismantling Procedures
PRESENTER: Ioan Doroftei

ABSTRACT. The mass electronics sector is one of the most critical generator of waste, both in terms of volume and materials content with harmful effects on the environment. The current amount of electronic systems is impressive while manual dismantling is a very common and non-efficient solution. Assuring an automatic procedure that can be replicated, is one of the task for efficient electronic waste recovery. This paper presents an architecture for advanced recovery of particular waste materials from computer and telecommunications equipment. The automated mechanical dismantling scheme is built on an eye-to-hand approach using a robotic system and a custom tool. The paper details the implementation layout and highlights the advantages of the proposed architecture.

Parameter estimation process for the dynamic model of robotic manipulators
PRESENTER: Dániel Szabó

ABSTRACT. This paper presents a method of estimating the dynamic parameters of robotic manipulators. The benefits of using the method called the modified Newton-Euler formula to determine the dynamic model of robotic arms are explained. It is shown, that the nonlinear model can be transformed into a linear-in-parameters model and the determination of the independently identifiable variables is explained. The differences between the applicability and efficiency of three estimators are presented, namely between the least-squares, weighted least-squares, and the maximum likelihood estimators. Several criteria are introduced to optimize the excitation trajectories. In a deterministic framework, the Frobenius norm as the condition number of the regression matrix can be used as the cost function to be minimized with a conditional nonlinear optimization algorithm, taken into account the limited joint angles, velocities, and accelerations. If the noises of the joint angles are not negligible, then the maximum likelihood estimator can be used in a stochastic framework and an iterative optimization can be applied to minimize the Cramér-Rao lower bound of the estimation.

Comparison of High-Precision GNSS systems for development of an autonomous localization system
PRESENTER: Huba Somogyi

ABSTRACT. In the near future the vehicles offering advanced driver assistance or fully autonomous operation. It will require increasingly accurate position information, available in all driving conditions and with 100 percent availability. In generally a single sensor is not able to meet these requirements alone and therefore it is necessary to use a combined sensor suite solution incorporating different kinds of sensors working together. As the only source of absolute position, velocity and time, Global Navigation Satellite Systems (GNSS) play a critical role in next generation positioning systems. In this paper we are dealing with modern GNSS solutions and presents the results of comparison of two high-precision GNSS systems in different driving conditions and presents conclusion for develop of autonomous localization system.

Two-Degrees-of-Freedom Controller Synthesis to Robustify the Computed Torque Method
PRESENTER: Barnabás Finta

ABSTRACT. This paper studies a novel controller structure for nonlinear systems with a robotic application. In robotics, the computed torque control method, which is roughly equivalent to state feedback linearization, is known for its high-speed tracking capabilities if the parameters of the dynamic model are known with precision. For MIMO linear systems, the robust controller synthesis technique is available to ensure robust stability and performance in the presence of model uncertainty. Two-degrees-of-freedom (2DOF) controllers also show good disturbance rejection properties. We propose a controller structure and design method where the feedback includes a nonlinear, linearizing component and a 2DOF robustifying component. One option available to control nonlinear systems is feedback linearization consisting of a feedback, and a nonlinear change of coordinates. The resulting closed-loop system is linear [1]. Let us recall that uncertainty in the model parameters may partly destroy the linearity of the closed-loop dynamics. For SISO systems, there is a necessary and sufficient condition for the existence of such a linearizing transformation. One may also use the formalism of Lie-Backlund equivalence [2][3] which states that if there exists a Lie-Backlund transformation that makes the system’s Cartan-field equivalent to the trivial vector field, then the system is differentially flat and can be linearized by dynamic state feedback. To deal with parameter or frequency domain uncertainty of MIMO linear systems, robust synthesis techniques can be used. By modelling a real physical system by an uncertain linear one and using robust control techniques to design an adequate controller, the stability and performance of the closed-loop can be guaranteed, despite model uncertainties, external noises and disturbances. The uncertainties are pulled out in a structured/unstructured uncertainty block with limited norm, and the closed-loop performance is specified by weighing transfer functions leading to a so-called augmented plant during the design phase. The robust controller is the solution of an optimization problem. We consider unstructured uncertainty and H-infinity optimization such that the solution is obtainable by numerical methods using Matlab [4]. This paper proposes a method to robustify nonlinear systems that are linearizable by state feedback. It is known that the exact linearization method to control nonlinear systems only achieves the nominal linear behaviour in closed-loop if the parameters of the system are precisely known. This uncertainty in the parameters leads to a closed-loop system, which may not be linear, or despite being linear, it has a different transfer function than the nominal model. Thus the closed-loop system’s performance and even its stability may be compromised. We aim to construct an outer loop to the exact linearized system, which can be synthetized using robust control theory. In order to ensure good disturbance rejection performance, we consider a 2DOF controller in the outer loop, which – together with the exact linearized inner loop – constitutes a cascaded structure. Recall that this is a modified structure compared to the earlier work presented in [5], where a serial robustifying controller was considered. The procedure of designing such a 2DOF robustifying controller is presented. The set of linear systems for the robust synthesis of the 2DOF controller is obtained by defining a sufficiently fine grid over the uncertain parameter space. For each vertice of this grid, the linearizing feedback calculated for the nominal set of parameters is applied, and the resulting closed-loop dynamics is approximated by the linear term in its Taylor series expansion. Tools such as Matlab support both the calculation of the uncertainty weighting functions and the robust synthesis of the 2DOF controller. Finally, the robust controller can be synthetized using standard H-infinity techniques. The performance of such a controller architecture is illustrated by a SISO and a MIMO example. The SISO example is a simple mass-spring system where the spring has nonlinear characteristics. All the necessary calculations are presented for this first example during the workflow for better understanding. The MIMO example is a 2DOF robot manipulator where the end effector moves in a plane. All the simulations with the new robustifying controller show significantly better performance than without it. We also analyze the difference in tracking performance, and disturbance attenuation between the serial compensator suggested in [5] and the proposed 2DOF controller.

15:00-16:40 Session 8B: Robotic applications and actuators
Location: Virtual Room B
A Novel Design of a Lightweight Magnetic Plate for a Delivery Drone
PRESENTER: Giuseppe Sutera

ABSTRACT. In recent years, drones have been widely used in multiple fields as well as to the military field for which they were born. The vertical flight capability makes these systems suitable for performing various tasks, in particular the delivery service. The delivery of things or basic goods in a very short time or in unreachable areas is a relevant application scenario. The systems used on the market adopt large electromagnets that affect the flight duration. Moreover, these devices have a high power consumption and heavy weight, thus reducing the maximum transportable payload. In this study, we propose a novel light magnetic plate composed by permanent magnets, that are able to pick and place any object as long as it has a ferromagnetic surface. This plate has been developed for the Mohamed Bin Zayed International Robotics Challenge 2020 (MBZIRC) [1], which is an international robotics competition for multi-robot systems. The challenge 2 envisaged a drone able to pick different types of bricks from the ground and to assemble them in order to build a wall according to an assigned pattern. The bricks had different colors, sizes and weights. The latter ranges from 1kg to 2kg and therefore the weight was the most relevant specification to be considered for the implementation on a drone. Several pick-and-place approaches adopt electromagnet [2], [3], however, two main issues arise with this solution. The first is the high energy absorption related to the transported object, as more energy is required to create a magnetic field, thus reducing time of flight of the drone. The second problem is the heavy weight of a typical electromagnet, which reduces the payload capacity. The idea is to change the classical approach moving on from an electromagnet to permanent magnets in order to extend the time of flight. For this reason our group developed a plate able to rise up to 2kg while reducing weight and power consumption. The gripper can pick and move ferromagnetic objects without consuming electric power and requires electric power only when the gripped object is released by a servomotor. The prototype was made using rapid prototyping techniques. All main components are 3D printed using ULTRAT material. The prototype made is 15 cm long, 10 cm wide and it has been designed in accordance with the specifications on the ferromagnetic surface size on top of the brick used in challenge 2. Finally we have developed and tested different vertical slide coupling systems, able to absorb the bump force between the plate and the drone. In fact, the grasping approaches with The authors are with the Dipartimento di Ingegneria Elettrica Elettronica e Informatica of the University of Catania - Viale A. Doria 6 - 95125, Catania (Italy) - Corresponding author: aerial vehicles is difficult because the contact between a drone and another object introduces a stability loss of the vehicle. The plate can be installed on both aerial and ground mobile platforms. The light weight and small size design guarantee an installation on a robotic arm, through a dedicated flange, which does not affect or limit the dexterity and postures of the manipulator. The on-field tests performed during the challenge trials proved the good performance of this magnetic plate compared to the solutions presented by other participants and the ability to move about 5 bricks during each trial.

Exoskeleton and lower limb prostheses: psychosocial aspects and acceptance of body image with the help of 3D printing

ABSTRACT. The new technologies that involve prostheses and exoskeletons have had a cutting-edge development in their mechanical, technological and functional nature, but studies on the subject involving the psychosocial impact of their users are not yet so advanced. It is observed that regarding the body image that the user has of himself, when he uses these mechanisms that complete his bodies, more studies are still needed. This article studies ways to achieve greater acceptance and higher self-esteem on the part of the patient with a lower amputation, who uses a transfemoral or transtibial prosthesis. The proposal was to introduce the use of custom 3D printed skins for prostheses provided by the Brazilian Unified Health System (SUS by its acronym in Portuguese). SUS, which serves the low-income population in Brazil, is responsible for 94% of all amputations performed in the country where the issue of aesthetics, self-esteem and acceptance of body image is even less common in this reality. The project proposal was to use a CAD (Computer Aided Design) and 3D Printing Software Technologies to create a prototype piece of Cover for Prosthetic Coating, with a design that highlights the beauty of the piece, with design and creativity in additive printing. Using a 3D scanner, a scan of a lower limb was performed from a dummy model, creating a "mesh" of polygons using 3D modeling software (3DMax) and then adhering this mesh to that member, exactly following the anatomical curvatures obtained in the exploration. After this process, this "mesh" was modeled, assigning the proposed design requests or customizations. Men and women wearing inferior prostheses participated, regardless of age. These contacts were obtained through the registration of the waiting list to attend the discipline or compulsory supervised internship of the Degree in Physiotherapy in orthopedics and traumatology at the Clínica Escola de Saúde do Centro Universitário Estácio de Santa Catarina, campus São José. Users were able to see a variety of designs with different proposals for texture, design, color and format. Finally, they gave their opinion on 3D printing as a form of visual identification. It was possible to evaluate the costs and understand that projects like these can have financial sustainability in "extension projects" that are an arm between society and the university, with students that produce improvements in society. As for the materials, plastic or polylactic acid (PLA) was used, which is an economic material and suitable for conceptual projects for prototypes, like this one, which do not have a mechanical function but only a coating. This material is suitable for printers selected here using Fused Deposition Modeling (FDM) technology. 3D technology combined with a differentiated design can provide a welcoming and social reintegration, minimizing depression and low esteem for people who carry robust equipment attached to them in the simplest daily tasks.

Master-Slave Control for a Pneumatically Actuated Low Pressure Soft Robotic Glove to Facilitate Bilateral Training for Stroke Patients
PRESENTER: Jaydip Desai

ABSTRACT. Stroke is one of the leading causes of disability worldwide. Most stroke survivors have some degree of paralysis immediately after stroke, which in many could be lost hand motor functions. Researchers have been working on improving existing rehabilitation techniques and hand exoskeleton designs to encourage a faster motor function recovery. However, many of these attempts seemed to be ineffective due to mechanical structure limitations of rigid hand exoskeletons. This paper focuses on designing a low pressure soft robotic glove to facilitate bilateral hand training. In this therapy method, a movement in the healthy hand generates a similar movement in the parietal hand. Researchers found that the activation of both cerebral hemispheres in rehabilitation exercises, improves outcomes significantly. To allow for this to happen, a design of soft actuators that can be quickly fabricated, and be operated on low pressures while achieving high flexion forces is proposed. The mechanical performance of the actuator was analyzed based on its blocked tip force capabilities and its capability in supporting full range of motion of the fingers. Then, a master-slave controller is proposed such that a glove (master) with flex sensors is worn on the healthy hand to generate a similar movement on the soft robotic glove that is worn on the parietal hand. The glove was successful in meeting the design requirements and assisting a healthy person in performing a pinch, tripod pinch, and full grasp movements.

A Medical AI Agent as a Tool for Neuropsychiatric Diagnoses

ABSTRACT. A Medical AI agent perceives its environment with sensors and aids diagnoses, communicating results to the physician and providing feedback to the patient. The agent augments existing clinician interaction by learning the evolving status of the patient over time. With limited contact between patients and physicians due to the onset of Covid-19, there is an increased need for telemedicine and the use of medical AI agents.

A neuropsychiatric diagnostic system utilizing a medical AI agent is proposed to capture physiological perturbations and patient responses in order to diagnose and track patients with psychiatric disorders like PTSD. Sensors are used to capture vital physiological parameters such as skin conductance, pulse rate, and blood oxygenation. These could be augmented with cameras and microphones that are used to record patient responses such as speech perturbations, oral responses, facial expressions, eye and pupil changes, and body movements to medical questions and stimuli.

Dependency of properties of a tactile sensor on the electrode design and conductive ink layer
PRESENTER: Jaromír Volf

ABSTRACT. The paper presents electrical characteristics of conductive ink of type Graphit 33, acting as a converter between applied pressure and output electrical signal. It mentions basic properties of the ink, including the explanation of the principle of function. Further it describes the measurement process on tactile sensors with circular electrodes of various types and dimensions. It studies the dependency of electrical resistance of the ink on the applied pressure. Various setups with different ink layer thicknesses are compared and their suitability to act as pressure converter is finally discussed.