Heterogeneous swarm complex multilevel reconfiguration based on Swarm Chemistry

Abstract

The need for the rapid and safe analysis of areas affected by emergencies is driving the search for innovative approaches in the field of autonomous systems. One such solution is the use of swarms of unmanned aerial vehicles to scan territories. This study presents a hybrid approach to the dynamic control of swarms of drones, combining self-organisation mechanisms based on ‘swarm chemistry’ with the optimisation of global parameters using an evolutionary algorithm. The proposed multi-level swarm reconfiguration module allows the behaviour of individual agents and the entire formation to adapt in response to environmental changes. A clustering module has been implemented to divide the scanning area into subregions, as well as a route planning system that considers the scanning width of each individual drone. To effectively distribute zones between agents, a proportional algorithm is employed that considers the functional capabilities of each device. Simulation results showed a reduction in mission execution time of more than twofold and an increase in the stability of swarm behaviour, even under conditions of heterogeneous composition and partial agent loss. This approach has significant application potential in emergency monitoring, victim search and damage assessment. This study formulates the architecture of a dynamic swarm reconfiguration system that can adapt to environmental changes in real time. This approach ensures the system’s flexibility and stability when performing tasks in difficult conditions. Further development of the model involves integrating machine learning methods to enhance adaptability and expanding to three-dimensional space to improve the accuracy and realism of simulations.