Publication
Linked Data as Medium for Stigmergy-based Optimization and Coordination
Torsten Spieldenner; Melvin Chelli
In: Hans-Georg Fill; Marten van Sinderen; Leszek A. Maciaszek (Hrsg.). Communications in Computer and Information Science (CCIS), Vol. 1622, Pages 1-23, Springer International Publishing, 7/2022.
Abstract
Optimization through coordination of processes in complex systems is a classic challenge in AI research. A specific class of algorithms takes for this inspiration from biology. Such bio-inspired algorithms achieve coordination and optimization by transferring, for example, concepts of communication in insect swarms to typical planner problems in the AI domain. Among those bio-inspired algorithms, an often used concept is the concept of stigmergy. In a stigmergic system, actions carried out by members of the swarm (or, in AI domains, by single agents), leave traces in the environment that subsequently work as incentive for following agents. While there is a noticable uptake of stigmergy as coordination mechanism in AI, we see the common understanding of one core element of stigmergic systems still lacking: The notion of the shared digital stigmergic medium, in which agents carry out their actions, and in which traces left by these actions manifest. Given that the medium is in literature considered the element ``that underlies the true power of stigmergy'', we believe that a well-defined, properly modelled, and technically sound digital medium is essential for correct, understandable, and transferable stigmergic algorithms. We therefore suggest the use of read-write Linked Data as underlying medium for decentralized stigmergic systems. We first derive a set of core requirements that we see crucial for stigmergic digital media from relevant literature. We then discuss read-write Linked Data as suitable choice by showing that it fulfills given the requirements. We conclude with two practical application examples from the domains of optimization and coordination respectively.