Multimedia
Abstract—In nature, when encountering unexpected uncertainty, animals tend to react quickly to ensure safety as the top priority, and gradually adapt to it based on recent valuable experience. We present a framework, namely EVOLVER, to mimic the bio-behavior for robotics to achieve rapid transient reaction ability and high-precision steady state performance simultaneously. In particular, the Koopman operator is leveraged to explore the latent structure of internal and external disturbances, which is subsequently utilized in an evolutionary model based disturbance observer to estimate the eventual disturbance. The resulting observer can guarantee a provable convergence in optimal conditions. Several practical considerations, including construction of a training dataset, data noise handling, and lifting functions selection, are elaborated in pursuit of the theoretical optimality in real applications. The lightweight nature of our framework enables online computation, even on a microprocessor (STM32F7 with 100 Hz control frequency). The framework is thoroughly evaluated by one simulation and three experiments. The experimental scenarios include: 1) trajectory prediction of an irregular free-flying object subject to aerodynamic drag, 2) indoor and outdoor agile flights of a quadrotor subject to wind gust, and 3) high-precision end-effector control of a manipulator subject to unwanted base moving disturbance. Comparison results show that the performance of our proposed EVOLVER is superior to several state-of-the-art model-based and learning-based schemes.
Supplementary code and data
For access to the code used in the chaotic Lorenz example visit here.
For access to the code and data used in the object trajectory prediction example visit here.
For access to the code used in the manipulator example visit here.
For details about guidelines for adjusting experimental parameters, the way to obtain the true value of the wind disturbance, and related work of chosen examples, visit here or here.
Update record
April-22, 2023 - First release.
May-18, 2023 - Update the video, in which the outdoor quadrotor experiment is supplemented.
Oct-23, 2023 - Update the Supplementary Materials.
Contant Email: jdjia@buaa.edu.cn, xiangyu_buaa@buaa.edu.cn