Sematectonic stigmergy: helping swarm robots find and co-manipulate the best object

Social insects are an abundant source of inspirations for swarm robotics thanks to their impressive cooperation which enables them to complete complex and difficult tasks. One of the very typical examples of such ability is found in the nest building activities of weaver ants where they collaborate to find and co-manipulate the softest part of a natural leaf to successfully curl it up. Inspired by this special feature of the weaver ant leaf-curling task, this thesis aims to study methodologies for swarm of simple robots to cooperate, to find and co-manipulate the most suitable object (the softest part of the leaf in this case) in complex working environments. The study started with the purpose of finding a suitable communication method and decision making algorithm for swarm robots to replicate the leaf-curling task of weaver ants. The study has shown that, without direct communication, using only indirect communication via changes of the working environment – the method called sematectonic stigmergy, simple robots can effectively collaborate to find and co-lift the softer edge. Playing the key role in achieving success for swarm robots is the decision making algorithm which can exploit the simple information that robots receive via sematectonic stigmergy. From the algorithm for the original robotic leaf-curling task, a more complete methodology was developed to help swarms of simple robots solve the generalized leaf curling task, in which the working environment is highly complex and completely unknown. The developed algorithms were investigated thoroughly using mathematical models and a comprehensive dynamic simulation model. The main advantages of each method were also verified by experiments with a group of physical robots on an artificial leaf.