Online Learning for Foot Contact Detection of Legged Robot Based on Data Stream Clustering.

Qingyu Liu,Yang Wang,Bing Yuan

doi:10.3389/fbioe.2021.771415

Qingyu Liu, Yang Wang + Show 1 more

Open Access

https://doi.org/10.3389/fbioe.2021.771415

Copy DOI

Abstract

Foot contact detection is critical for legged robot running control using state machine, in which the controller uses different control modules in the leg flight phase and landing phase. This paper presents an online learning framework to improve the rapidity of foot contact detection in legged robot running. In this framework, the Gaussian mixture model with three sub-components is adopted to learn the contact data vectors corresponding to running on flat ground, running upstairs, and running downstairs. An online data stream learning algorithm is used to update the model. To deal with the difficulty in obtaining contact data at landing moment online, a “trace back” module is designed to trace back the contact data in the memory stack until the data meet with the probability contact criterion. To test if the foot is in contact with the ground, a projection method is proposed. The acquiring data vector during the leg flight phase is projected onto an independent random vector space, and the contact event is triggered if all projected random variables fall within 1.5σ of the corresponding Gaussian distribution. Experiments on a legged robot show that the presented algorithm can predict the foot contact 16 ms in advance compared with the prediction using only leg force, which will ease the controller design and enhance the stability of legged robot control.

Highlights

The ability to negotiate unstructured terrain is the most significant advantages of the legged robot compared with wheeled and tracked vehicles
As we only focus on verifying the foot contact detection, a classical finite state machine is adopted as a high-level control scheme
When the contact event is triggered by the Gaussian mixture model (GMM) prediction module or leg force criterion, the update module will trace back the contact data in the memory stack until the data meet with the probability criterion

Summary

Introduction

The ability to negotiate unstructured terrain is the most significant advantages of the legged robot compared with wheeled and tracked vehicles. Due to the discrete foot point characteristic in legged locomotion, like the human and other legged animals, the robot goes through a series of foot contact in locomotion. Based on different foot contact states, a finite state machine is usually adopted to identify the gait phases, and different control modules will plan the leg motion trajectory to balance the robot. Robust perception of the foot contact arises as a crucial ability in legged robot control. Though a force sensor mounted on the foot could be a straightforward solution (Wagner et al, 2017), it is damaged due to the foot–ground impact and the unknown rough terrain. The foot force sensor would increase the inertia of the leg

Methods

Results

Discussion

Conclusion