Grasp Pose Detection and Feature Extraction Using FHK-GPD and Global Average Pooling in Robotic Pick-and-Place Systems

To overcome challenges including occlusions, intricate object geometries, and dynamic environments, robotic pick-and-place systems in industrial and logistical applications need sophisticated grip position recognition. This work merges GAP and Feature Hierarchical Kinetics-Based Grasp Pose Detection, FHK-GPD in order to increase robustness and predictive capabilities. The methods integrate FHK-GPD with GAP to globally extract and analyze the hierarchical features so as not to overfit. The result of the proposed method achieved 98.9 % accuracy, 96.4 % precision, and 97.1 % F1 score. In conclusion, FHK-GPD and GAP improve grasp recognition, thus providing industrial robots with robustness and efficiency. Optimization will be the main focus of future effort.