InterJournal Complex Systems, 542
Status: Accepted
Manuscript Number: [542]
Submission Date: 20502
Revised On: 20502
Amorphous Predictive Nets
Author(s): Michael Howard ,David Payton ,Regina Estkowski

Subject(s): CX.14, CX.13

Category: Article


This paper describes our approaches for coordinating the actions of extremely large numbers of distributed, loosely connected, embedded computing elements. In such networks, centralized control and information processing is impractical. If control and processing can be decentralized, the communications bottleneck is removed and the system becomes more robust. Since conventional computing paradigms provide limited insight into such decentralized control, we look to biology for inspiration. Due to progress in the miniaturization of sensors and computing elements and in the development of necessary power sources, large arrays of networked wireless sensor elements may soon be realizable. The challenge is to develop software that enables such amorphous arrays to self-organize in ways that enable the sensing capabilities of the whole to exceed that of any individual sensor. Our goal is to devise local rules of interaction that cause useful computational structures to emerge out of an array of distributed sensor nodes. These distributed logical structures appear in the form of local differences in sensor node state. These local state differences serve to form distributed circuits among nodes, allowing groups of nodes to perform cooperative sensing and computing functions that are not possible at any single node. Further, since the local differences emerge and are not pre-programmed, there is never a need to assign specific functions to specific nodes. In this paper we describe two methods, each using only local interactions between nodes to detect the presence and heading of some local transient property of the environment (e.g., presence of a warm body). These methods provide a purely distributed means of computing the direction and likely destination of a sensed movement, with no need for centralized data analysis or explicit data fusion. Such a prediction could activate sensors ahead of the movement of the sensed object, turning on more expensive sensing functions that are normally dormant to save power. An active minefield could use the techniques to attract mines to the most likely avenue of approach. Streetlights could be turned on ahead of cars on a road less traveled.

Retrieve Manuscript
Retrieve Previous Revision's Abstract
Submit referee report/comment

Public Comments: