A Binocular Active Vision System Using Space Variant Sensors: Exploiting Autonomous Behaviors for Sp

Future space programs will take advantage of advanced robotics; visual capabilities, in particular, could play a very important role, either to provide robots with independent behaviors and to improve human remote control. The paper describes a prototype of a binocular vision system based on space variant sensing technologies. The system is aimed to operate autonomously implementing a set of basic behaviors such as the ability to track oating objects and to change the focus of attention in relation to visual information coming from the periphery of the visual eld. These capabilities could be helpful to support outer space operations as well as autonomous remote teleoperations. Particular attention has been devoted to the description of the system architecture, the technological sensing peculiarities and the image processing algorithms implemented so far. Many applications, like services in hazardous areas, space operations or harsh environments exploration would beneet of robotic support. For instance, the inspection of some outer part of a spacecraft or a satellite (may be an antenna dish). clearly involves sensorial capabilities (recognition, hand-eye coordination, tactile sensing) but, most importantly, the ability to operate safely in outer space environment. This skill requires a very long training and consequently high costs. Therefore it seems reasonable, and, to a great extent, cost eeective, to think of autonomous robot systems able to operate in outer space performing speciic tasks of maintenance or surveillance. A diierent context in which autonomy could be a key issue is that of remote space teleoperations. The ability to see what happens in the remote environment is an essential requirement for a system operator, but experiences in ground-to-space as well as space-to-space communications show that the presence of transmission delays make any continuous control of a remote system very diicult 15]. A possible way to diminish the role played by delays is to provide remote systems with simple but basic autonomous capabilities. For example, an active vision system could monitor the operational space of a robot arm, providing procedures to avoid local collisions and to manage unexpected events. But, what are the features really important in developing this kind of systems? We know that human formidable skills are hard to replace: these limits are even more evident when trying to build intelligent machines, able to reproduce even simple behaviors. The main requirements we can foresee for these machines are robustness, real-time responsiveness, continuous control of the variables involved in the process. …