Target allocation method, device and system based on multi-target deep reinforcement learning

• Main Authors: LIU JUNTAO, WANG JUNZE, GAO ZIWEN, YAO DI
• Format: Patent
• Language: Chinese; English
• Published: 19.07.2024
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; PHYSICS

The invention relates to a target allocation method, device and system based on multi-target deep reinforcement learning. The method mainly comprises the following steps: inputting multi-target allocation problem instance information and a weight vector set; matrix processing is carried out on the multi-target allocation problem instance information, an encoder module of the strategy model is input, and an output result of the encoder module is obtained; calculating a weight coefficient of the decoder module; calculating a Pareto solution; and traversing each weight vector, repeatedly calculating the weight coefficient of the decoder module and calculating the Pareto solution according to the traversed weight vector, obtaining a Pareto solution set of the multi-target allocation problem instance, and completing the solution. According to the method, the Pareto solution set solving of the multi-target allocation problem can be realized. 本发明涉及一种基于多目标深度强化学习的目标分配方法、装置及系统。其方法部分主要包括：输入多目标分配问题实例信息，以及权重向量集合；对多目标分配问题实