Time Series Motion Generation Considering Long Short-Term Motion

Abstract

Various adaptive abilities are required for robots interacting with humans in daily life. It is difficult to design adaptive algorithms manually; however, by using end-to-end machine learning, labor can be saved during the design process. In our previous research, a task requiring force adjustment was achieved through imitation learning that considered position and force information using a four-channel bilateral control. Unfortunately, tasks that include long-term (slow) motion are still challenging. Furthermore, during system identification, there is a method known as the multi-decimation (MD) identification method. It separates lower and higher frequencies, and then identifies the parameters characterized at each frequency. Therefore, we proposed utilizing machine learning to take advantage of the MD method to infer short-term and long-term (high and low frequency, respectively) motion. In this paper, long-term motion tasks such as writing a letter using a pen fixed on a robot are discussed. We found differences in suitable sampling periods between position and force information. The validity of the proposed method was then experimentally verified, showing the importance of long-term inference with adequate sampling periods.

Imitation Learning for High Speed General Object Manipulation

“…Using the proposed method, the operating speed is determined based on the peak frequency calculated using the fast Fourier transform (FFT), and the slave responses are concatenated and inputted into an NN. Variable speed operation is achieved by simply incorporating the operating frequency as an input, even though the method is almost identical to that described in [16] [17]. The proposed method can be regarded as a combination of imitation learning with parametric biases, in which the physical parameters of robotic motions can be adjusted [26].…”

An Independently Learnable Hierarchical Model for Bilateral Control-Based Imitation Learning Applications

“…In fact, past research has shown that extracting movement primitives by utilizing the interaction force results in better generalization abilities [27]. Subsequently, we proposed bilateral control-based imitation learning that enables robots to execute tasks requiring force adjustment and fast behavior [28]- [32]. Bilateral control is a remote-control technique for leader and follower robots with force feedback [31].…”

A New Autoregressive Neural Network Model with Command Compensation for Imitation Learning Based on Bilateral Control

“…We propose bilateral control-based imitation learning as a method of using force information [14]- [16]. Bilateral control is a teleoperation method where a human operates the master robot and the slave robot performs tasks in the workspace [17]- [20].…”

“…In addition, by introducing bilateral control to imitation learning, the master measures the action force, and the slave measures the reaction force during the data collection phase. We used bilateral control for data collection, and the method showed effectiveness in tasks that required force adjustment and demonstrated fast motion [14]- [16]. In bilateral control-based imitation learning, an S2M model that predicts the next master state from the current slave state was used.…”