A Load Distribution Estimation Method for Multiple Trucks Incorporating Attention Mechanism

Accurately predicting the load distribution within the truck hopper during automated loading processes is crucial for ensuring operational efficiency and safety. Existing studies in this domain are predominantly limited to single devices and single types of materials. Moreover, they typically rely on the availability of complete information, which significantly restricts their applicability in real-world scenarios where such information is often incomplete or partially available. To address these limitations, we propose a novel multitruck load prediction network—mask mass R-CNN (MM R-CNN). This is the first network designed to simultaneously estimate the load distribution across multiple trucks, providing a more comprehensive and practical solution. The proposed MM R-CNN extends the mask R-CNN by incorporating a mass head branch. This addition allows the network to filter and select appropriate prediction boxes under specific conditions to capture regions of interest, thereby enabling simultaneous detection and estimation for multiple trucks. Furthermore, we have designed an attention fusion architecture that significantly enhances the accuracy of MM R-CNN, especially when dealing with incomplete target information. Experimental results demonstrate that the trained model achieves an impressive accuracy of over 95%, highlighting its potential for practical deployment in industrial settings. We believe our method will provide a new direction for improving material load distribution estimation.