.png){kind=icon}
.png){kind=icon}
Return
Spray Coating Coverage Path Planning Based on Multi-layer Feature Aggregation RainbowNet
.png){kind=icon}
PRE
AI
DOI：10.1016/j.measurement.2025.117710.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Automated putty spraying robot for high-speed rail body shell addresses the inefficiency and inconsistency inherent in manual processes. The crux of this automation lies in path planning, which must optimize coverage while minimizing redundancy. Insufficient coverage results in inadequate putty application, whereas excessive repetition leads to prolonged spray gun inactivity, potentially causing nozzle obstruction and reduced efficiency. Conventional coverage path planning methods often fail to meet requirements. So, we propose a novel approach based on deep reinforcement learning: the Multi-layer Feature Aggregation Rainbow Net (MFARainbowNet). This method incorporates several key components: MFANet for processing environmental features, Traceback mechanism to balance exploration rewards and Action Mask to guide action selection. Experimental results demonstrate the efficacy of our proposed method in achieving comprehensive coverage. Compared to existing approaches, our method achieves a remarkably low redundancy rate of 0.5 %. Moreover, it exhibits robust performance in complex environments, maintaining a redundancy rate of 7.22 %.
Keywords:
Putty Application
Coverage Path Planning
Deep Learning
Reinforcement Learning
Journal
IF:
5.6
Papers:
2.0W
Citations:
5.4W
Organization
No organization information available