SemanticMediaWiki As A Collaboration Platform For Self Organizing Agents

From semantic-mediawiki.org
Jump to: navigation, search
SMWCon Fall 2017SemanticMediaWiki As A Collaboration Platform For Self Organizing Agents
SMWCon Fall 2017
SemanticMediaWiki As A Collaboration Platform For Self-Organizing Agents
Talk details
Description: This talk by Nicole Merkle presents how Semantic MediaWiki can be used by self-organizing agents as a knowledge exchange platform.
Speaker(s): Matthias Frank
Type: Talk
Audience: Everyone
Event start: 2017/10/06 13:55:00
Event finish: 2017/10/06 14:20:00
Length: 25 minutes
Video: not available
Keywords: SWRL, SPARQL, Reinforcement Learning, Agent-based systems
Give feedback

Motivation

Nowadays software agents play a crucial role for supporting people in solving heterogeneous tasks. This brings many challenges for the developers of agent-based systems. Tasks are domain specific and vary due to the objectives of a certain domain. For every solvable task, an agent requires to be programmed somehow. This can be difficult and cumbersome as the environments in which agents are acting are complex and in the most cases stochastic. In order to deal with these problems, an agent has to be flexible regarding context-changes and the characteristics of an environment. For this reason, an agent has to be able to observe environmental states (e.g. sensed by IoT devices) and develop a strategy (policy) as well as interpret the context in order to decide the required actions to perform.

In this talk, Nicole Merkle shows how Semantic MediaWiki can help to address the mentioned challenges by serving as a collaboration platform for self-organizing, autonomous agents in order to enable them to share, rate and re-use their autonomous learned strategies for solving varying and context-dependent tasks. Moreover,this talk presents a Semantic MediaWiki extension Semantic StateChart that simplifies for domain experts or developers the provision of semantic task descriptions which are processed by appropriate agents. The presented approach leads-regarding the programming of agents-to a shift from how-to-do-something to what-to-do. The developer is no longer responsible for the development of strategies but the agent itself, empowered by machine learning and semantic web technologies.

Topics

The following aspects will be considered in this talk:

  • The concepts of using SMW as a collaboration platform for autonomous agents.
  • The concepts of self-programming agents by means of semantic Web technologies and reinforcement learning.
  • Presentation of the Semantic StateChart extension.