A three-dimensional space line network facility intelligent planning method based on deep reinforcement learning
By constructing a three-dimensional spatial grid decision-making environment and defining a multi-objective reward function, and combining deep reinforcement learning and curriculum learning mechanisms, a proximal policy optimization algorithm was designed. This solved the problem of coordinating multiple constraints and economic objectives in three-dimensional spatial network facility planning, generating efficient and superior facility layout schemes, and improving the scientific and economical nature of engineering decisions.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- INST OF DEFENSE ENG ACADEMY OF MILITARY SCI PLA CHINA
- Filing Date
- 2026-03-31
- Publication Date
- 2026-07-17
AI Technical Summary
Existing methods struggle to effectively coordinate multiple engineering constraints and economic objectives in 3D spatial network planning, leading to solutions falling into local optima. Furthermore, deep reinforcement learning suffers from training convergence difficulties and poor constraint compliance when dealing with complex engineering rules.
A deep reinforcement learning-based intelligent planning method is adopted to construct a three-dimensional spatial grid decision-making environment, define a multi-objective reward function, establish technical constraints, and design a proximal policy optimization algorithm based on a curriculum learning mechanism. The algorithm training process is optimized through an Actor-Critic network architecture and an importance sampling mechanism to generate a network facility layout scheme that satisfies multiple engineering constraints and is economically optimal.
It enables the automatic generation of facility layout schemes with excellent comprehensive technical and economic performance in complex 3D environments, improves global optimization capabilities and training efficiency, and promotes the digital and intelligent development of infrastructure planning.
Smart Images
Figure CN122413629A_ABST