MOVIE: a Velocity-aided IMU Attitude Estimator for Observing and Controlling Multiple Deformations on Legged Robots

Matthieu Vigne (1, 2) , Antonio El Khoury (1) , Marine Pétriaux (1) , Florent Di Meglio (2) , Nicolas Petit (2)

1 Wandercraft
2 CAS - Centre Automatique et Systèmes

This paper presents a novel observer for estimating the state of a legged robot with deformations in the structure, taking into account the dynamic nature of walking motions. Specifically, it considers a robot with several punctual deformations, whose rotations are estimated using low-cost IMUs. To compensate for the linear acceleration corrupting the measurement of gravity in the accelerometer, this observer exploits the robot's kinematics model. Namely, it combines gyroscope and accelerometer signals, encoder measurements, and attitude estimates from the previous IMUs to reconstruct the linear velocity of a given IMU. This reconstructed measurement is used to estimate the tilt (roll and pitch) of the current IMU, which is then exploited to propagate the linear velocity to the next sensor. The resulting cascade of observers, called MOVIE, is experimentally benchmarked on the exoskeleton Atalante, where it outperforms our previous observer, relying on a quasistatic assumption. This observer is then used in closed-loop control, to compensate for the effect of deformations while walking, resulting in an improved walk.

Domaines: Robotique [cs.RO]

Matthieu Vigne, Antonio El Khoury, Marine Pétriaux, Florent Di Meglio, Nicolas Petit. MOVIE: a Velocity-aided IMU Attitude Estimator for Observing and Controlling Multiple Deformations on Legged Robots. IEEE Robotics and Automation Letters, 2022, 7 (2), pp.3969-3976. ⟨10.1109/LRA.2022.3149025⟩. ⟨hal-03511198⟩