Classification and Management of Computational Resources of Robotic Swarms and the Overcoming of their Constraints

Swarm robotics is a relatively new and multidisciplinary research field with many potential applications (e.g., collective exploration or precision agriculture). Nevertheless, it has not been able to transition from the academic environment to the real world. While there are many potential reasons, one reason is that many robots are designed to be relatively simple, which often results in reduced communication and computation capabilities. However, the investigation of such limitations has largely been overlooked. This thesis looks into one such constraint, the computational constraint of swarm robots (i.e., swarm robotics platform). To achieve this, this work first proposes a computational index that quantifies computational resources. Based on the computational index, a quantitative study of 5273 devices shows that swarm robots provide fewer resources than many other robots or devices. In the next step, an operating system with a novel dual-execution model is proposed, and it has been shown that it outperforms the two other robotic system software. Moreover, results show that the choice of system software determines the computational overhead and, therefore, how many resources are available to robotic software. As communication can be a key aspect of a robot's behaviour, this work demonstrates the modelling, implementing, and studying of an optical communication system with a novel dynamic detector. Its detector improves the quality of service by orders of magnitude (i.e., makes the communication more reliable). In addition, this work investigates general communication properties, such as scalability or the effects of mobility, and provides recommendations for the use of such optical communication systems for swarm robotics. Finally, an approach is shown by which computational constraints of individual robots can be overcome by distributing data and processing across multiple robots.