Research and development of digital objects for robotic simulators, including dynamic human models
Project number: №24-29-00564
Head of the project: professor Evgeni Magid
About the project
Robotic systems are not only expensive objects that interact with people and the environment, but also objects that become dangerous if they are not handled properly. Therefore, before new robotic interaction protocols and algorithms can be integrated into a real robotic system, they are thoroughly tested, firstly in the laboratory, and then, to the extent possible, in a real target environment. In addition to testing in a real environment, virtual ones in simulators are actively used in the robotic testing. Currently, the new approach virtual testing in simulators has become an integral part of the process of creating new technologies in the field of robotics, which is used at the initial stages of development and saves a great number of man-hours and financial costs when creating the first robotics system software and hardware prototypes.
The vast majority of digital models of real robotic systems are created by the manufacturers of these systems or their associated scientific groups in universities and organizations. However, during the robotic projects testing, especially in the field of robot-human interaction, it is important for the simulator to include not only robotic systems high-quality digital models, but also digital environmental objects and digital models of people performing various actions in this environment. Unfortunately, today in the popular simulators’ virtual environments for robotic systems there is a rather limited amount of digital environmental objects of a high quality. For simple small-sized objects (e.g. furniture), for which a simple software implementation of a digital model is sufficient, the quantity and quality of such models in the most popular simulators (e.g. Gazebo or Webots) is satisfactory, however, more complex dynamic models of digital objects are either completely absent and the user has to create them himself every time for the particular project (e.g. doors opening) or must be content with primitive low-quality models (e.g. Gazebo's standard human model).
The project is aimed at qualitative improvements in the field of modelling and virtual validation of robotic systems prototypes, which will allow more realistic and efficient modelling of interaction protocols and algorithms for such systems in the virtual environment of popular robotic simulators. As part of the proposed project, the digital models of the environment available in the most popular simulators, including human models, will be evaluated. The most urgently needed complex digital objects for generic virtual experiments will be identified and the possibilities of their creation in popular simulators will be assessed. Scientific novelty lies in: development of specific requirements for the models of the proposed digital objects and their functionality, which will ensure their realism and efficiency; development of criteria for assessing the quality of the model; development of experimental methods for assessing the quality of the model; digital objects modelling in accordance with the proposed requirements. The digital object model libraries and graphical user interfaces developed within the project are of great practical value, they will help the user create their own models based on the models created by the Kazan Federal University (KFU) team, as well as customize the ready-made models taken from the existing libraries.
During the first year of the project, the following work was carried out:
1) A literature review on the stated topic was carried out, aimed at analyzing and systematizing knowledge about digital objects, robotic simulators and digital human models.
2) Popular robotic simulators used for mathematical and physical modeling of robots were analyzed. Special attention was paid to standard digital objects used in virtual environments. The capabilities and limitations of existing simulators were studied in the context of providing ready-made digital objects for modeling.
3) The functional capabilities of popular robotic simulators were analyzed. Based on the characteristics of the simulators and their compliance with the requirements for digital models of humans and robots, Gazebo and Webots were selected as the most promising robotic simulators.
4) Two key types of complex digital objects were selected: a digital human model and a model of the omnidirectional mobile robot ArtBul (developed earlier with the participation of the project team) (Fig. 1).
Fig. 1. Digital model of the mobile robot "ArtBul", developed with the participation of the project team members.
5) The existing digital human models used in robotic simulators were tested and analyzed, their advantages and disadvantages were identified, and their ability to simulate physical, cognitive and visual parameters was evaluated.
6) Functional requirements, aimed at providing the integration of digital objects with simulators, the realism of the movement of digital objects, interaction with the environment and control, were developed. Non-functional requirements for digital objects for robotic simulators were developed.
7) For a structured approach to the analysis and development of digital model architecture, a classification is proposed that includes aspects such as visual structure, functionality, visual and collision detailing, scope of application, control methods, and degrees of interactivity. To evaluate the quality of digital human models in robotic simulators, a testing methodology was developed aimed at checking the characteristics of the models.
8) A modular architecture of digital objects was developed. It ensures flexibility in the application of the developed digital models and simplifies the integration of additional components.
9) Complex digital human models were developed for Gazebo and Webots. 20 pilot models were created, 10 for each of the two simulators (examples are shown in Fig. 2-3).
Fig. 2. The first pilot versions of digital objects in Gazebo, created within the project.
Fig. 3. Pilot versions of digital objects in Webots.
10) Five virtual worlds were created and adapted for the experiments. To evaluate the models, 13 test scenarios were developed, aimed at checking the functionality, optimization and compliance with the requirements of the simulators.
11) Based on the intermediate results obtained in the first year, 4 publications were published and indexed in the Scopus or Web of Science databases.
Publications
[1] Sulaiman S., Pillai A., Eryomin A., Tsoy T., Magid E. Kinematic Modelling, Workspace Analysis and Static Structural Analysis of a 3 DoF Manipulator for an Elevator Automation // The 4th International Conference on Computer, Control and Robotics (ICCCR 2024). – 2024. – pp. 300-304.
[2] Sulaiman S., Eryomin A., Tsoy T., Magid E. Methodology of Design and Modelling Projects for Robotics and Mechatronics Course of Intelligent Robotics Master Program // 2024 IEEE 7th Eurasian Conference on Educational Innovation (IEEE ECEI 2024). – 2024. – pp. 410-413.
[3] Abbyasov B., Zagirov A., Gamberov T., Li H., Magid E. Vision-based autonomous navigation for medical examination using a UR3e manipulator // International Conference on Artificial Life and Robotics (ICAROB 2024). – 2024. – Vol.29. – pp. 308-311.
[4] Celiker B., Sulaiman S., Tsoy T. A Design of a Modular Mobile Robot for Rescue Operations // International Conference on Artificial Life and Robotics (ICAROB 2024). – 2024. – Vol. 29. – pp. 267-271.
