Recently, there has been a fairly rapid increase in interest in the use of UAV swarms both in civilian and military operations. This is mainly due to relatively low cost, greater flexibility, and increasing efficiency of swarms themselves. However, in order to efficiently operate a swarm of UAVs, it is necessary to address the various autonomous behaviors of its constituent elements, to achieve cooperation and suitability to complex scenarios. In order to do so, a novel method for modeling UAV swarm missions and determining behavior for the swarm elements was developed. The proposed method is based on bigraphs with tracking for modeling different tasks and agents activities related to the UAV swarm mission. The key finding of the study is the algorithm for determining all possible behavior policies for swarm elements achieving the objective of the mission within certain assumptions. The design method is scalable, highly automated, and problem-agnostic, which allows to incorporate it in solving different kinds of swarm tasks. Additionally, it separates the mission modeling stage from behavior determining thus allowing new algorithms to be used in the future. Two simulation case studies are presented to demonstrate how the design process deals with typical aspects of a UAV swarm mission.
Widespread access to low-cost, high computing power allows for increased computerization of everyday life. However, high-performance computers alone cannot meet the demands of systems such as the Internet of Things or multi-agent robotic systems. For this reason, modern design methods are needed to develop new and extend existing projects. Because of high interest in this subject, many methodologies for designing the aforementioned systems have been developed. None of them, however, can be considered the default one to which others are compared to. Any useful methodology must provide some tools, versatility, and capability to verify its results. This paper presents an algorithm for verifying the correctness of multi-agent systems modeled as tracking bigraphical reactive systems and checking whether a behavior policy for the agents meets non-functional requirements. Memory complexity of methods used to construct behavior policies is also discussed, and a few ways to reduce it are proposed. Detailed examples of algorithm usage have been presented involving non-functional requirements regarding time and safety of behavior policy execution.
26 September 2019 in Annals of Computer Science and Information Systems