Route planning for multiple surveillance autonomous drones using a discrete firefly algorithm and a Bayesian optimization method

This paper focuses on the planning of routes for autonomous drones used for the surveillance of a given area. Routes consist of permutations of checkpoints selected by a human supervisor. The permutations are evaluated by a cost function that models the uncertainty of the monitored area’s situation, according to the elapsed time since the last visit of the different checkpoints. This paper suggests to address the problem of permutation optimization by employing a discrete version of the firefly algorithm. It presents three exploration strategies: swapping random pairs of elements, shuffling a random subset of elements, and moving toward a noisy firefly. In addition, it also suggests to use a Bayesian optimization technique for tuning the firefly algorithm hyper-parameters, instead of commonly used grid search techniques.