Robot Manipulation Strategy for Explosives Disposal in Cluttered Scenes

Compared to human operator, robotic arm possesses the advantages of high accuracy and controllability in handling and manipulating hazardous materials. Due to the complexity and variability of application scenarios, such as the objects are stacked, it is a great challenge to accurately select target objects and manipulation mode in the fields of security and demolition. To address these problems, robotic arm systems require capabilities in target object detection, object relationship reasoning, and manipulation strategy selection. In this paper, we propose a framework with the abilities of reasoning about the unknown relationships between target objects and interfering objects, predicting manipulation sequences, and adopting appropriate manipulation strategies for hazardous object detection and manipulation processing. We conduct a series of experiments in indoor single-target, multi-target, and outdoor scenarios to complete the assigned tasks with improved manipulation sequences. These results serve as evidence of the effectiveness of the proposed framework.