Ant Colony Algorithms andits applications to Autonomous Agents Systems

Abstract: With the latest advancements in autonomous agents systems and technology, there is a growing interest in developing control algorithms and methods to coordinate large numbers of robotic entities. Following this line of work, the use of biologically inspired algorithms based on swarm emerging behaviour presents some really interesting properties for controlling multiple agents. They rely on very simple instructions and communications to develop a coordinated structure in the system. Particularly, this master thesis focuses on the study of Ant Colony algorithms based on stigmergy interaction to coordinate agents and perform a certain task. The first part focuses on the theoretical background and algorithm convergence proof, while the second part consists of experimental simulations and results. For this, some metric parameters have been developed and found to be especially useful in the study of a simple path planning test case. The main concept developed in this work is an adaptation of Shannon Entropy that measures uniformity and order in the system and the weighted graph. This parameter has been used to study the performance and results of an autonomous agent system based on Ant Colony algorithms. Finally, this control algorithm has been modified to develop an event-triggered control scheme. Using the properties of the weighted graph (Entropy) and the sensing of the agents, a decentralized event-triggered method has been implemented and tested, and has been found to increase efficiency in the usage of system resources.