DIANA: Distributed and Safe Autonomous Navigation for a Swarm of Autonomous Vehicles

The autonomous systems are typical examples of complex distributed cyber-physical systems (CPS). The main characteristics of such systems is the use of autonomous vehicles. They are increasingly used in various mission-critical tasks such as surveillance and rescue operations. To execute the required tasks, the autonomous swarm systems should fulfil such important dependability requirements as safety and reliability. On the hand, while designing a swarm system, we should guarantee that the robotic systems in the swarm do not collide with each other and objects in the operating environment, i.e., ensure motion safety. On the other hand, to ensure that the robots have sufficient resources to reliably complete the required goals, we should also achieve efficiency while implementing the swarm mission, i.e., minimise the travelling distance of the robots vehicles. In this thesis, we propose a novel integrated approach that ensures motion safety and efficiency while planning and controlling an operation of swarms of autonomous robots. We validate our approach in different case studies and compare them with some state-of-the art benchmarks. Moreover, we rely on formal modelling to derive the safety constraints guaranteeing that the swarm system can cope with both predicted and dynamically emerging safety hazards. We define an architecture of the controlling software that combines static and dynamic mechanisms for safe and efficient swarm control and navigation. To ensure efficiency, while preserving safety, we propose a new parallel algorithm for swarm mission planning. This algorithm is a combination of evolutionary computing methods, machine learning and deterministic approaches that coordinated by a central management component. The algorithm controls the swarm actions on three different layers: the offline, online and vehicle layer as well as allows us to plan and optimise at the run-time the routes of the vehicles to maximise safety while minimising the travelling distance. Our solution promotes a holistic approach to designing CPS – from a formal requirements definition to a software implementation that fulfils the defined requirements in an efficient way. The results of benchmarking demonstrate that our approach allows safe and efficient control of swarms.