• This paper addresses the challenge of target allocation under temporal constraints in scenarios involving multi-weapon coordination against multiple target types. It over-comes technical limitations of traditional planning methods, such as handling sequential decision-making, dynamic responses, and coupling relationships, by designing a dynamic target allocation framework based on a dueling deep Q-network with target compression and segmented decision-making. The main contributions of this paper can be summarized as follows: • A multidimensional state representation integrating ammunition inventory, target type, and damage status is designed, encoding time-dependent damage mechanisms as state transition rules. • The allocation process is modelled as a multi-stage decision problem, simultaneously generating the target allocation matrix and strike sequence through action mapping. • A dynamic target priority compression mechanism is introduced to reduce problem dimensionality and enhance strategy stability. This paper addresses the challenge of target allocation under temporal constraints in scenarios involving multi-weapon coordination against multiple target types. It overcomes technical limitations of traditional planning methods, such as handling sequential decision-making, dynamic responses, and coupling relationships, by designing a dynamic target allocation framework based on a dueling deep Q-network with target compression and segmented decision-making. First, for a typical combat scenario in which a ground-based unmanned formation engages three types of targets using two kinds of artillery shells, a damage mechanism model incorporating temporal dependencies is constructed. Second, using a multi-dimensional state space to encode both munitions inventory and target states, a segmented decision mechanism is established to achieve joint optimization of the allocation matrix and strike sequence. Finally, a target compression module is introduced to resolve strategy stability issues within the high-dimensional action space. Simulation experiments demonstrate that the proposed method significantly improves strike effectiveness while satisfying temporal constraints, offering a novel technical pathway for intelligent target allocation.
Yao et al. (Wed,) studied this question.