Modeling and Consensus-based Control of Multi-Agent Systems

Nowadays, Unmanned Aerial Vehicles (UAVs) applications are crucial in several domains, either civilian or military as these systems are mainly used within hostile environments while ensuring user integrity or industrial and academic sectors while coping with specific needs and services. For instance, UAVS are significantly used in the field research missions, as for example volcanic-activity’s monitoring, climate change studies, crowds-management, management/surveillance and epidemiological tracking. Multi-Agent Systems (MASs) represent a trending topic in the UAVs field since the related literature shows that many applications are enhanced. Indeed, MAS is an ensemble of individual agents sharing some information to achieve a collective objective, as the cooperation could enhance the effectiveness of the mission accomplishment with respect to the action of a single agent. Collective behaviour implies an efficient inter agent’s and surrounding environment information flow (topology) and a coordination protocol must shape the formation, the agreement value and the synchronization strategy for completing the mission. Many examples can be found in nature, for example in migratory birds, forming a network of collaborating entities that fly together in different shaped coordinated formations (flocking) or other species that agree about an interest variable (motion, communication, ...) to enhance mutual benefits, that include predators’ survival, collective foraging, long distance migration, and, thus, the acquirement of more dynamic efficiency as the main advantage. In particular, the flocking behaviour is a one of the key aspects of the MASs development and implementation, including the interaction and communication between the elements of a group, the perception that each entity has of the whole flock (or sub-group of the flock) and of each flock mate, the detection of external disturbances and obstacles. Despite the evolving trending topic and the promising MAS applications, numerous technological and scientific challenges are still to be addressed and solved: firstly the efficient definition of a communication protocol, the implementation on real UAVs, subjected to several software and hardware failures. The main focus of this thesis is the MAS model definition for consensus problems. In detail, by efficiently representing the system by graphs, the consensus algorithms are verified in MATLAB&Simulink environment and completed including the knowledge of the leader of the desired trajectory. A centralized-based model is firstly proposed, but a distributed model is the target of the work. Different formations are compared (e.g. spatial placement, sensors’ sensitivity and detection area, addition of obstacles, …) in terms of observability and controllability, aiming to the identification of the topologies that allow to reconstruct the initial states by implementing an observer and, afterwards, a PID controller. At last, the model are experimented on real MASs together with the initialization of a new environment for UAVs wherein basic flight tasks will be tested. In the future, the model will be expanded by considering topologies changing overtime, such as the addition/removal of nodes or by considering the efficient action of the formation when avoiding an obstacle or going through a narrow passage and by an in-depth application to the real environment, with the option of estimating external disturbances.