The purpose of the project is to realize a new information infrastracture called PII:the perceptual information infrastracture, which consists of multiple perceptual agents and computer networks . The perceptual agents have ability of perception, computation, and communication to make the real world accessible from computers. The PII maintains information of the real world obtained by the perceptual agents in order to support cognitive behaviors.
As the first step toward realization of the PII, we investigate network design of vision agents (perceptual agents having vision sensors), localization of vision agents, correspondence problem of multiply observed objects by agents, and modeling global behavior of robots and humans.
We have developed a prototype of the distributed vision system (DVS) as an actual example of the PII, and have studied new technologies being necessary to build the PII.
The prototype consists of a 1/12 scaled small town model, sixteen vision agents and two mobile robots. On the prototype system, we showed that the DVS can navigate the robots robustly even in complicated and dynamic environment . Figure 20 shows the prototype town model and the DVS, in which realistic shadow and road texture are presented so that the prototype can be used as a testbed to verify practical system problems.
Figure 20: Distributed Vision System
Figure 21 shows scenes obtained by sixteen vision agents which is navigating two robots. Frames with light gray and dark gray rims shows the vision agents observing robot A and B respectively. The robots can dynamically select arbitarary agents having information suitable for the navigation.
Figure 21: A Robot Navigated by Vision Agents
We have also developed a real size prototype of the DVS in Figure 22. This prototype consists of vision agents equiped with omnidirectional vision sensors. Because these sensors have very wide view angles, the system can track human behaviors robustly in real time [4,5].
Concepts and fundamental problems of the PII are considered as follows:
As the first issue, we have already developed technologies to make omnidirectional sensors small and low cost. As the second issue, a robuts algorithm to calculate qualitative relations of agents' positions have been developed [3,6]. The other issues will be future works.
Figure 22: Distributed Vision for Tracking Human Behaviors
So far our interests have been on behavior recognition of humans and supporting robot behaviors. We will proceed to the next issues (3 - 6) and go toward our final goal which is to establish design principles of the PII.
The PII covers natural environment protection, support of cultural and economic activity of humans in scale of the earth. We consider the PII as the next generation of global information infrastractures.
Servers TD> A>