Enhanced neighborhood search with knowledge transfer for multi-aircraft collaborative task allocation in emergency scenarios

Purpose: This paper aims to propose a multi-start collaborative variable neighborhood search (VNS) method, named MCVNS, which effectively reuses and transfers informative knowledge among optimization cycles, to address the multi-aircraft collaborative task allocation in emergency scenarios. Design/methodology/approach: The authors design a multi-start initialization strategy, using four distinct constructive heuristics to generate a variety of starting points for the VNS optimizers. Moreover, the authors design seven problem-specific local search operators tailored to enhance the effectiveness of these optimizers. To facilitate full collaboration among the VNS optimizers starting from different points, the authors develop a novel collaborative knowledge transfer strategy that efficiently reuses promising neighborhood sequences across different optimizers. Findings: Comprehensive experimental studies on the public benchmark instances indicate that the proposed MCVNS significantly outperforms the state-of-the-art methods. Further analysis verifies the effectiveness of different devised components within MCVNS. Originality/value: This paper demonstrates that reusing well-performing neighborhood sequences, previously explored by the VNS optimizers during the optimization process, can effectively guide subsequent optimization cycles and bring possible benefits.