Flapping-wing robots mimic the flight patterns of natural flying creatures, granting them natural camouflage. They hold promising applications in the military reconnaissance, urban surveillance, and various other fields [1–3]. However, the current limitations in size often restrict the suitability of existing bird-like flapping-wing robots, such as E-Flap [4] and USTBird [5], for indoor tasks. Furthermore, while micro flapping-wing robots such as Delfly and H2Bird excel in navigating confined spaces, their progress towards attaining stable autonomous flight capabilities requires further development [5]. Accurately following preplanned paths is essential for enabling micro flapping-wing robots to autonomously execute tasks. Among the path following algorithms currently in use, the vector field algorithm stands out for its ability to minimize cross-track error and control effort in guiding miniature air vehicles [6], but its application to micro flapping-wing robots has yet to be explored. This work presents two primary contributions. First, a novel X-wing flapping-wing robot capable of hovering and stable flight was designed. Second, a vector field-based guidance algorithm was proposed for the autonomous straight-line path following of the robot, and flight experiments verified the effectiveness of the control system.

Design of the experimental platform. The micro X-wing flapping-wing robot we designed, depicted in Figure 1 (a), boasts a wingspan of 27 cm and a weight of 18 g. A micro motor, connected to a gear set, drives the wings to flap periodically. For turning control, the robot employs a micro servo to manipulate the T-shaped tail. Additionally, a communication module interfaces with the flight control board, enabling real-time connectivity with the upper computer. The robot’s head and tail are equipped with five reflective markers, aiding recognition by the motion capture system, as depicted in Figure 1 (b). This system swiftly captures position and quaternion data from the robot in motion, achieving a delay of less than two milliseconds. Subsequently, this data is transmitted to the upper computer for additional processing as shown in Figure 1 (c).