Community Research and Development Information Service - CORDIS

A robot which keeps its eyes open

Realistic 3 dimensional (3-D) models are becoming increasingly important in many computer applications. A mobile robot prototype promises a sense of "being there" for telepresence and virtual reality applications.
A robot which keeps its eyes open
Achieving a 3-D model of our surroundings has now been made possible by advances in computer science. Attempts have been made to record the spatial and visual complexity of real-world environments, in two distinct steps. Firstly, by creating a graphical 3-D representation of the environment under study, and then by "painting" it (texture mapping) to give the best possible approximation to the natural visual appearance.

At the Joint Research Centre's (JRC) Institute for Systems, Informatics & Safety (ISIS), an integrated approach using a mobile robot prototype has been used. The robot known as AEST (Autonomous Environmental Sensor for Telepresence) can tour the inside of a building and automatically create a 3-D map of the interior complete with surface texture information.

The AEST is an autonomous mobile platform carrying a specially designed scanning head. The sensor head is composed of a laser range finder with a rotating mirror for capturing 3-D structures and a video camera for capturing texture information. The embedded software carries out the registration and integration of data from different viewpoints and the registration of video texture, and builds the final 3-D reconstructed models.

The platform contains separate navigation and reconstruction computers and has twenty-four ultrasonic sensors that assist with collision avoidance. The size of the unit is comparable to that of a person to ensure that it can be taken to all the places where people are likely to pause when looking around a building. It also supports views at eye level to better convey the feeling of being present at the real scene.

AEST has the ability to capture spatial data at different points by navigating automatically between locations. By combining what is seen from neighbouring capture positions, surfaces that would be occluded (hidden) from one position are recorded. The system also detects occlusions and works out where the optimum location of the next viewing point should be.

The viewer can select the location from which to see the environment, unaware of the capture points from which the 3-D model was constructed. This ability to construct such models automatically from real structures opens up a range of opportunities, in areas such as real-estate, design, games and simulators.

Related information

Record Number: 80815 / Last updated on: 2005-09-18
Domain: IT, Telecommunications