Adaptive conflict resolution for multi-UAV 4D routes optimization using stochastic fractal search algorithm

Abstract

The increasing unmanned aircraft system (UAS) applications in urban environments pose challenges for safe and efficient low altitude air traffic management. As an essential enabler to meet these challenges, pre-flight 4D routes optimization is required to conduct conflict detection and resolution (CD&R) and to generate conflict-free flight routes before departure. Existing studies on strategic deconfliction cover several types of strategies such as scheduling or rerouting. However, a single type of strategy used to solve different types of conflicts may lead to an unsafe and inefficient way of conflict resolution. This paper proposes an adaptive decision-making framework to optimize the resolution strategies used for different types of conflicts with explainable mechanisms. The proposed framework is formulated as a double-layer optimization problem with the considerations of scheduling, speed adjustment, and rerouting strategies for conflict resolution. The first layer of the framework is established as a probabilistic selection model to make decisions on which strategy should be selected for what type of conflict. The second layer is developed as a mixed-integer nonlinear programming (MINLP) model to optimize the decision variables of the strategies selected by the first layer. To solve the proposed double-layer optimization problem, we introduce and improve a novel meta-heuristic stochastic fractal search (SFS) algorithm with two major improvements of a penalty-guided fitness function and an exploitation-exploration balancing scheme. Simulation results demonstrate that the proposed adaptive conflict resolution framework successfully optimizes the strategies used for each type of flight conflict, which subsequently optimizes the 4D routes with significant reductions in total operational cost, number of flight conflicts, and flight delays. The improved stochastic fractal search (ISFS) algorithm is also proved effective and reliable in solving the proposed optimization problem in different traffic density scenarios.