Sensor Fusion-Based Perception Methods for Robotic Arms

Robotic arms are expanding from structured industrial scenarios to dynamic unstructured operation scenarios, which puts forward higher requirements for perception capabilities. A single sensor is difficult to meet the demands of robustness, real-time performance and completeness in complex operations. However, multi-sensor fusion technology, by integrating multi-modal information such as vision, force and touch, can significantly enhance the environmental perception and adaptability of robotic arms. This paper systematically reviews the classification and typical configuration strategies of robotic arm sensors, and focuses on analyzing the principles, applications and effects of multi-sensor fusion algorithms at different control levels. Finally, it summarizes the current challenges and looks forward to the future development direction, with the aim of providing theoretical references for related research.