Quadrotor-based multimodal hopping and flying locomotion significantly improves efficiency and operation time as compared to purely flying systems. However, effective control necessitates continuous estimation of the vertical states, as thrust (insufficient for flight) in the aerial phase creates non-ballistic behavior. A single hopping continuous state estimator has been shown (Kang 2024), in which two vertical states (position, acceleration) are measured and velocity is estimated through a technique requiring multiple sensors (IMU, lidar, depth camera, contact force sensor), and computationally intensive calculations (12-core, 5 GHz processor), for a maximum hop height of ~0.6 m at 3.65 kg. This poses a significant challenge to the development of light-weight, high-performance, low observable, jamming and electronic interference resistant hopping systems; especially in perceptually degraded environments (e.g., dust, smoke). Here we show a trained Kalman filter based hopping vertical state estimator (HVSE), requiring only vertical acceleration measurements. The training uses hopping data from the robot and a motion capture system to adapt a general framework to the specific system; including high impact behaviors. Our results show the HVSE can estimate more states (position, velocity) with 32% of the mean-absolute-percentage-error in the hop apex height (height error/ground truth) of the next best inertial navigation technique (12.5%), running ~4.2x faster (840 Hz) on a substantially less powerful processor (dual-core 240 MHz) with over ~6.7x the hopping height (4.02 m) at 20% of the mass (672 g). The presented general HVSE, and training procedure make the methodology broadly applicable to other robots.
翻译:暂无翻译