Android-type robots have always attracted the human mind, being the most expressive result of his desire for creativity. Currently, research and development in the field of artificial intelligence is ahead of its application on real humanoid robots. Meanwhile, many research groups of students and postgraduates are inspired by the idea of applying their developments to a physical model. Due to the complexity of manufacturing an Android-type robot, such projects usually fade away at the level of implementing mechanics and adapting the original idea to the limitations that have arisen in the process of working on mechanics. The article describes an anthropomorphic robot based on the experience of creating the "Greta" robot. Examples of training tasks are shown.
The purpose of this work is to study the possibilities of introducing a universal electronic-mechanical platform for creating robots and project activities in the field of cyber-physical systems and artificial intelligence into the educational process. 1.1
In 2012, students of Physics and Mathematics Lyceum #239, under the guidance of a
group of teachers and specialists from scientific organizations, created a robot of the
Android type "Greta" [
The following tasks were set:
creating manipulator hands that can interact with human hands through the
touch of electromagnetic sensors,
implementation of computer vision elements for recognizing human hands,
implementation of speech interface elements,
creating a positive appearance that is attractive even for a child,
different game modes: teacher, student, partner.
The Greta robot won gold medals at the WRO world robot Olympiad in Malaysia [
Results of the First Stage: the set goals were achieved with minimal means, implemented the necessary control algorithms, the system is optimized for Autonomous operation without an operator, the Greta robot has worked together for about 2 months at various exhibitions and festivals, there is a need to continue at a qualitatively new level, realized the need for a production base.
The success of the first version of the project inspired its creators to search for new
solutions. There were already a number of anthropomorphic robots for various purposes
in the world. Large corporations and universities create the most impressive models, so
it makes sense to consider budget projects as analogues that should be based on the
experience of creating them. For the AR601M Android robot produced in Russia, a
control algorithm has been developed since 2014 [
Spontaneous movements of robot manipulators, similar to human hands, are presented
in the development of Japanese laboratories, which was called Alter [
In 2017, the Greta project was joined by a commercial company [
The developers were assigned the following tasks: maintaining and improving the functionality of manipulator hand movement and contact with human hands, moving around the room on the Omni platform, ability to modify manipulators for actions with objects, functional elements of the universal promo robot, using computer vision and machine learning to recognize human hands and other objects, voice interface.
Within a few iterations, a ready-to-use launch platform was created, providing opportunities for the development of all tasks. Currently roboblather has a number of advantages compared with the previous version of "Greta": consists of a small set of standardized parts, involves modular Assembly on nodes, as well as rapid change of shells, parts, and tools, provides quick replacement of modules with improved ones, equipped with a modular mobile platform on Omni-wheels with height adjustment, added additional sensors and indications for non-verbal communication (for example, increased the number of contact pads on the palm, built-in led matrix in the face), the modules are hand made on the basis of digital servos with feedback on position, speed, current and load.
Now, the Greta 2.0 robot platform consists of two independent parts powered by batteries.
Torso: two reinforced arm modules (6 degrees of freedom per arm), a head module (2 degrees of freedom); the torso sensor is represented by capacitive palms, web cameras, and a 3D sensor. Mobile platform: the pelvis module on a support bearing (1 degree of freedom), three linear actuators for adjusting heights, four motors with Omniwheels and encoders; the sensor system is represented by a gyroscope-accelerometer in the pelvic part, ultrasonic rangefinders, in the future it is planned to install a laser scanner.
The roboplatform based on high behavioral and design resource. In the future, it will be possible to create multi-variant designs for a wide range of tasks on the basis of easily replaceable modules, connectors and motors with a feedback system. 4
Thus, a universal technical platform has been created that allows connecting various
teams and teams that want to contribute to the development of robotics. It is also
expected that this platform will be compatible with multiple electronic components, as
well as with open resources on the Internet. On the example of a robot platform, training
in programming complex kinematic structures with nonlinear behavior, a control
system for navigation in a room, and the use of neural networks for self-learning a robot
when interacting with a person and the environment is assumed. The roboplatform is
universal from the point of view of management, suitable for learning programming at
different levels of complexity, since it can be controlled both with AVR
microcontrollers and with TRIK robotic controllers [
Examples of training tasks that can be solved using the Greta 2.0 robot. Development of a system for determining the position of human hands from images of palms using a multi-layer neural network. Study of the possibility of using a neural network to develop a system for determining the optimal trajectories of movements of the Android robot's limbs that repeat the movements of human hands. Development of a behavior model for an Android robot playing hand clapping game with a human using a neural network. Development of a system for synchronizing the movements of an Android robot that comes into tactile contact with a human, and research on the possibility of using a neural network to determine points of contact.
Creating multi-agent robotic systems.
The considered robot platform has a high starting potential for inclusion in the work of educational and scientific groups. The Greta 2.0 robot can be successfully used for training in the areas of mechatronics and robotics, industrial cyber-physical systems, etc.