Collaborative sensors data processing and environment information infrastructure as means to support autonomous actions of service robots

• Authors: Vitaliy Rybak
• Venue: SENSIG '10/MATERIALS'10 Proceedings of the 3rd WSEAS international conference on Advances in sensors, signals and materials
• Year: 2010

The new trends in robotics research have a general goal of developing personal and professional service robots that presupposes the robots operation in an unstructured environment. In this case one of the key problems is the problem of constructing in real time the model of dynamic robot surroundings using sensors data processing. Different sensors are differed markedly between each other by the price and by their technical characteristics such as accuracy of measurements, spatial resolution, speed of measurements and time of data processing. The more information about surroundings it is possible to get, the more expensive and time consuming is the sensor. To solve existing basic contradictions, we suggest to use a combination of various ways, both traditional - working out of more perfect and inexpensive sensors and data processing means, and no conventional - such which, at first sight, can seem not concerning a considered problem. Offered working out of a canonic stereo system on the base of commercial cameras and the simplified control system of a choice of the calibrated focal length of monocameras can serve as an example of the traditional way. Collaborative processing of various data, visual and non-visual, and information infrastructure of robot environment relate to the no conventional ways. Proposed collaborative sensors data processing is used to cut down analyzed space of scene by directed selection of the area of interest and to replace the pattern recognition problem solving by the verification of the state of known scene. As more often used sensors do not manage to be synchronized, the scheme of sensors data labeling by the measurement instants is offered. Designing the robot environment information infrastructure, we follow the practice of human being in creation of the information infrastructures to provide his safe existence. We propose visual information landmarks that can be easily recognized by the mono- and stereovision systems and permit them to define their 3D spatial position with respect to landmark coordinate system. It allows us to realize an environment model representation as a hybrid of topological and metrical maps.