UAV path planning in presence of occlusions as noisy combinatorial multi-objective optimisation

A realistic noisy combinatorial problem on surveillance by unmanned aerial vehicle (UAV) in presence of weather factors is defined. The presence of cloud coverage is considered as a posterior Gaussian noise in the visibility region of the UAV. Recent studies indicate that recombination-based search mechanisms are helpful in solving noisy combinatorial problems. The search strategy of univariate marginal distribution algorithm that includes only selection and recombination, which has a close association with genepool crossover, proves to be beneficial in solving constrained and multi-objective combinatorial problems in presence of noise. This paper proposes a solution methodology based on multi-objective UMDA (moUMDA) with diversification mechanisms for the multi-objective problem of UAV surveillance. To obtain a well-spread set of Pareto optimal solutions, relevant diversification mechanisms are important. Numerical simulations show that moUMDA with and without K-means clustering provides better quality solutions and a more diverse Pareto optimal set than NSGA-II in solving this noisy problem.