Controlled Flight of an Insect-Scale Flapping-Wing Robot via Integrated Onboard Sensing and Computation

Abstract

Aerial insects can effortlessly navigate dense vegetation, whereas similarly sized aerial robots typically depend on offboard sensors and computation to maintain stable flight. This disparity restricts insect-scale robots to operation within motion capture environments, substantially limiting their applicability to tasks such as search-and-rescue and precision agriculture. In this work, we present a 1.29-gram aerial robot capable of hovering and tracking trajectories with solely onboard sensing and computation. The combination of a sensor suite, estimators, and a low-level controller achieved centimeter-scale positional flight accuracy. Additionally, we developed a hierarchical controller in which a human operator provides high-level commands to direct the robot's motion. In a 30-second flight experiment conducted outside a motion capture system, the robot avoided obstacles and ultimately landed on a sunflower. This level of sensing and computational autonomy represents a significant advancement for the aerial microrobotics community, further opening opportunities to explore onboard planning and power autonomy.

Figures (7)

• Figure 1: Overview of robot design, flight package selection, and sensory autonomous flight. A) An image illustrating the robot lands on a sunflower. B) An image of the robot without integrating sensors. C) The three PCB boards carry an IMU (top), an MCU (center), and an optical flow and a ToF (bottom), respectively. The model numbers are reported in grey text. D-E) Top (D) and bottom (E) views of the 1.29-g robot with the integrated sensors and MCU. F) Lift force comparison shows the scaled-up robot module (red) achieves 46% lift increase compared to a prior design (blue). G) A side view composite image showing a flight performed outside a motion capture arena. The robot avoids obstacles and lands on a sunflower. The background is blurred to remove facial information of the experimenter. Scale bars in (A-E) and (G) represent 1 cm.
• Figure 2: Onboard sensing and state estimation. A) A high-level schematic of the sensing, estimation, and control pipeline. B) A schematic of the sensing and cascaded estimator design. A complementary filter and two observers sequentially estimate robot attitude, altitude, and lateral position, respectively. C) Evaluation of state estimation accuracy. The robot conducts a 4-s flight under feedback control with offboard motion capture data. The onboard measured and estimated position, velocity, attitude, and angular velocity are compared against the offboard motion capture data.
• Figure 3: Hovering demonstration with onboard sensing. A) A composite image sequence of a 12-s hovering flight that is 5 cm above ground. B) Comparison of position measurements based on either onboard estimation or offboard motion capture. C) Measured position data of five repeated experiments. In (B-C), the three columns correspond to the x, y, and z positions, respectively. D-F) Composite image sequence, position comparison, and repeated experiments of 12-s hovering flights set 10 cm above ground. Yellow dots in (A) and (C) represent the hovering setpoint. Scale bars in (A) and (C) represent 1 cm.
• Figure 4: Trajectory tracking demonstrations with onboard sensing. A) A side-view composite image sequence of a 6-s setpoint switching flight. B) Desired, estimated, and measured robot position corresponding to the experiment in (A). C) A side-view composite image in which the robot tracks an “infinity” sign. D) Desired, estimated, and measured robot position corresponding to the experiment in (C). E) A top-view composite image in which the robot tracks a planar circle. F) Desired, estimated, and measured robot position corresponding to the experiment in (E). Scale bars in (A), (C), and (E) represent 1 cm.
• Figure 5: Pollination demonstration. A) A schematic of a hierarchical controller that integrates high-level human issued positional commands and low-level position tracking with onboard sensing. B) A top-view composite image sequence of a 30-s flight demonstration. C) Side-view image sequence that illustrates robot takeoff, obstacle traversal, flight altitude control, and landing. D) Desired, estimated, and measured robot positions corresponding to (B-C). E) A top-view composite image sequence of a 15-s flight demonstration. F) Desired,estimated, and measured robot positions corresponding to (E). In (D) and (F), the grey regions represent portion of the flights where motion tracking is available. Motion capture became unavailable when the robot flew out of the capture flight volume.