Robot Tartalo, jamba

2008/06/01 Ochoa de Eribe Agirre, Alaitz - Elhuyar Zientziaren Komunikazioa Iturria: Elhuyar aldizkaria

• Robotika
• Adimen artifiziala
• Unibertsitateak

The people who work on the third floor of the Computer Science Faculty of San Sebastian consider it normal to meet Tartalo in the corridors. You find yourself but not. In fact, this smart metre and a half length machine avoids all the obstacles found along the way thanks to sensors installed around your 'body': sonars ultrasonic detectors, infrared sensors and lasers. The latter, for example, measure the distance at which everything within the 180 degree angle is located. UPV researchers have not manufactured the robot, the University has bought it and the research team is developing its capabilities.

With these sensors, and with the computer that is the brain of the robot, Tartalo would only have to move from one side to the other safely, that is, a basic English navigation called wandering. However, this team of researchers from the Department of Computer Science and Artificial Intelligence wants the robot to be able to attend a specific ordered point.

Guiding inside the buildings

The best-known machines that will drive us from a specific point to a destination are GPS navigation systems. However, they do not work within buildings and it would not be practical to create a database that collects the blueprints of all buildings in the world. Therefore, the team of Basilio Sierra is based on biomimetic systems for the development of Tartalo. That is, the robot does the same thing that people and animals would guide in a new place: explore the place and take reference points. But in order for a machine to do what living beings do with their intuition, computer scientists must insert a series of data and calculation programs.

The buildings are semi-structured environments, in all of them you can find common areas and Tartalo has taught you to know four of them: room, corridor, hall or entrance and crosses. Thus, if we took the robot to our house, it would first perform a process of self-location, giving a weight lap to learn the location of these spaces. With this process, the machine would form a kind of topological map, and we should only show the name of each of the places it learned. To do this, UPV researchers are designing human-robot interaction systems. For example, a voice recognition system and a touch interface are being adapted for the robot to understand the commands.

Single eye, sharp vision

To identify what's ahead -- for example, to separate a room from a hallway -- Tartalo uses the only eye that gives it its name: its camera. It measures the images taken through this camera, compares them to its database and is based on the probability of deciding what else those images look like. For example, the robot knows that if an area is long and narrow it is a corridor.

The most important thing they have shown Tartalo is to identify the doors. In fact, to reach most of the indicated places, the robot must access through a door. Therefore, it has the camera located at the same height as the door knob, which will help you identify the door. At this time the system is programmed to search and pass through doors when circulating through a corridor. If the door is closed, as arms have not yet been installed, the door is hit by the 'legs', with two or three collisions.

The team of researchers of the UPV-EHU aims to develop the navigation system of the robot, and to achieve this it is essential to detect the doors. From there, Tartalo will have to learn to distinguish many other things, such as faces, voices or anything you are asked to bring. But each of these actions involves the development of a specific program, which is currently outside the research line of the Group of Robotics and Autonomous Systems. However, this robot will gradually incorporate the capabilities developed by other research groups.