Scientific goals

WP1: The scientific goals of this package is research of the use of machine learning in cloud-based robotics with emphasis on research of text processing algorithms. The text processing will be examined within the emotional HRI and also an improvement of the HRI using cloud resources. Results of the work will be presented at international conferences and journals with impact factor. 

Expected contribution: The expected contribution is the study a development of systems, which will be accessible in the form of cloud service and will be able to analyze an emotion of a speech expressed by a text. This task includes text processing and analyzing with the aim to identify expressed emotions. For the purpose of text processing and analyzing, the research will be focused on the methods of symbolic artificial intelligence. It is not excluded that will emerge new methods as a result of current methods modification. The systems, which will be developed in this work package, will be able to deploy in the social interaction human-robot (HRI). A robot will recognize an expressed emotion and will select an adequate motion, for highlighting the human emotion. This way the robot can be part of an interaction between humans. From the mentioned information we can deduce the goal of usage multimodality in the improvement of human-robot interaction. It is necessary to highlight that, the system developed in this work package, will be a part of the integral software. Results will be scientific reports, papers, part of the integral software.

WP2: The scientific goals of this package is novel object recognition from an image and it will form a separate module in an integrated software. In this module the artificial intelligence techniques used for image processing, feature extraction from static images and also suitably adapted classifiers based on neural networks will be implemented. The goal is to revise the classification methods able to work with a dynamic feature space, i.e. Results will be new algorithms and also software module for object recognition from an image publications at scientific conferences and journals. 

Expected contribution: The expected contribution is the study and a development of a cloud-based system, for purposes of processing static and dynamical images (images and video) gained using sensors installed in an intelligent space and also from robotic systems. Phrase 'image processing' means, the extraction of the image features and consequent processing using the way, allowing human face recognition, human emotions, recognition of number of people in the room etc. From the artificial intelligence point of view, we will focus on classification methods, mainly on ARTMAP and Deep learning neural networks, featured with deictic learning in the cloud robotics. Results will be scientific reports, papers, part of the integral software.

WP3: The scientific goals of this package is the study of the Intelligent Space which should will serve as a support for robots during cooperation with each other, or HRI. Similarly, the cloud will serve as a support for the robotic systems during task execution in the form of data and processing. The AI will focus mainly on the data obtained from the sensors so the robotic platform can use these data as efficiently as possible. The results will be the publications at conferences and journals.

Expected contribution:  The expected contribution is the design a prototype of an intelligent space and to continuously create the so-called Robot Ready Environment. The intelligent space and its data collection will be realized in the form of a hierarchical database. Simultaneously, we will researching to define the notion "Ambient Robot" for realization of selected tasks, e.g. navigation, action planning, obstacle avoidance, interpretation of map information about the intelligent space. The goal is the normalization of communication between a robot and its environment, mutually among robots for needs of cooperative robotics. As methods of artificial intelligence the following methods will be studied and developed: reinforcement learning, evolutionary approaches (particle swarm optimization, migration algorithms), fuzzy and neuro-fuzzy systems.

WP4: The scientific goals of this package is to design and implementation a novel algorithm and software for creation of robot behavior in social HRI. To achieve the goal, mainly learning methods based on reinforcement will be used extended by other means of subsymbolic artificial intelligence. Another goal is to replace the teleoperator in the human-robot interaction with a semi-autonomic robot behavior. In both cases cloud computations will be utilized. The work results will be presented at international conferences and scientific journals. 

Expected contribution:  The expected contribution is the study and development of a system providing individual teaching experience while keeping track with the users using human-robot interaction without any physical contact. This is a complex research goal which entails many social human-robot interaction aspects (SHRI). The reinforcement learning method will be used for examination of various approaches dedicated for incremental learning and stepping up from teleoperation presence to a semi-autonomous (autonomous) robot behavior/presence. The main challenge will probably be designing the algorithms for particular human-robot interaction scenarios. Results will be scientific reports, papers, part of the integral software.