The invention relates to the technical field of two-dimensional rectangular part nesting, in particular to an intelligent two-dimensional rectangular part nesting method based on deep
reinforcement learning, which comprises the following steps of: converting a two-dimensional rectangular part nesting problem into a combined optimization
decision problem containing a target function and a constraint condition, constructing a deep
reinforcement learning framework by adopting a PPO (Point-to-Point-Oriented)
algorithm, and obtaining a two-dimensional rectangular part nesting result; and part positioning based on NFP and BL fusion. According to the two-dimensional rectangular part intelligent nesting method based on deep
reinforcement learning, autonomous learning is realized through a deep reinforcement learning framework, a strategy is optimized from historical experience, repeated calculation is avoided, performance is continuously improved, quality and efficiency are solved in a balanced mode through a PPO
algorithm and a pointer network, a large-scale problem scheme is output in a second level, the generalization ability is high, and the method is suitable for popularization and application. The method adapts to tasks of different scales and shapes, a multi-dimensional reward function flexibly responds to an optimization target,
trepanning is regarded as a sequence decision, and the material
utilization rate is effectively improved.