Abstract:
Protecting borders from the illicit transfer of people, weapons, goods, and UAVs is essential for
a country’s security, economic strength, and territorial integrity. Conventionally, most borders
deploy human surveillance which entails high costs for training and supplies, making security
personnel vulnerable due to direct exposure to the threats. Utilizing the advancement in tech nology, a move towards technology-assisted border surveillance is the need of time to help reduce
the extra cost and loss of human life while improving surveillance quality. In this regard, we pro pose a technology-assisted solution for border surveillance by employing a swarm of unmanned
aerial vehicles (UAVs) to improve surveillance quality. This study presents a framework for
the optimal placement of UAVs in a 3D environment to provide maximum coverage for cer tain border surveillance tasks, such as image capturing, intruder identification, and search and
rescue. For analysis, we apply several meta-heuristic algorithms including particle swarm opti mization (PSO) and genetic algorithms (GA) to model this scenario. Our research demonstrated
that these algorithms underperform in border surveillance tasks as compared to other scenar ios. This implies that the constraints provided by border surveillance, such as cost efficiency,
near-perfect accuracy, and scale invariance, require a better optimization method for obtaining
desirable results. To address this, we propose a hybrid technique combining PSO properties with
selected features of GA operations to incorporate diversity in the solution and to avoid local
optima, local search was used, and adaptive parameter settings were used for dynamically ad justing hyper-parameters. For experiments, we assess 4 different area types: urban, sub-urban,
high-rise urban, and dense urban. The hybrid technique outperformed the PSO and GA when
examined in a 3d 100m sub-urban area. The algorithm was able to cover 95% of the aerial and
ground plane with only 2 UAVs. In comparison, PSO required more than 9 UAVs to obtain 90%
coverage, while GA required at least 6 UAVs. Simulation results show that the GLP-Hybrid
algorithm proved invaluable in all 24 test cases with varying area sizes and area types.
