Integrating swarm intelligence with bayesian networks for continuous UAV-based surveillance in dynamic environments

The loss of wildlife due to illegal poaching activity poses threats on both the survival of iconic animal species and the livelihood of local communities. This research proposes a distributed surveillance model in which a UAV swarm autonomously coordinates continuous surveillance in a dynamic environment. The adaptive behaviour of poachers has the potential to negatively affect surveillance performance and is therefore taken into consideration through the proposed ACOSG model. The novelties of this research are twofold. A mission selection algorithm is proposed that addresses the deficiencies of the existing HAPF-ACO model while improving on surveillance effectiveness. Bayesian learning is applied to dynamically prioritise surveillance efforts of the proposed HAPF-BLACOPS model. Additionally, the learning rate of both poachers and UAVs is analysed to determine whether surveillance remains effective in response to adaptive poacher behaviour. Simulation results show that the proposed model significantly outperforms the current stateof-the-art HAPF-ACO model. Prioritisation of surveillance efforts is achieved through the use of (A)BNs, such that coverage of the target area is reduced by 30%, while maintaining the surveillance effectiveness of the current state-of-the-art. It is found that the interactions between the APF and ACO modules limit the extent to which the (A)BNs’ predictions influence the UAVs’ spatiotemporal patterns and therefore limit the effects of prioritisation on surveillance performance as well. The adaptive capabilities of the poachers and the learning rate of the UAV swarm do not significantly affect surveillance performance and the loss of wildlife, due to a limited amount of newly gained experience. Future research opportunities are identified that can improve the influence of the (A)BN on surveillance performance and prioritisation.Aerospace Engineerin