Superpixel Segmentation Based Synthetic Classifications with Clear Boundary Information for a Legged Robot.

Yaguang Zhu,Chao Ma,Qiong Liu,Kailu Luo,Bo Jin

doi:10.3390/s18092808

Yaguang Zhu, Chao Ma + Show 3 more

Open Access

https://doi.org/10.3390/s18092808

Copy DOI

Abstract

In view of terrain classification of the autonomous multi-legged walking robots, two synthetic classification methods for terrain classification, Simple Linear Iterative Clustering based Support Vector Machine (SLIC-SVM) and Simple Linear Iterative Clustering based SegNet (SLIC-SegNet), are proposed. SLIC-SVM is proposed to solve the problem that the SVM can only output a single terrain label and fails to identify the mixed terrain. The SLIC-SegNet single-input multi-output terrain classification model is derived to improve the applicability of the terrain classifier. Since terrain classification results of high quality for legged robot use are hard to gain, the SLIC-SegNet obtains the satisfied information without too much effort. A series of experiments on regular terrain, irregular terrain and mixed terrain were conducted to present that both superpixel segmentation based synthetic classification methods can supply reliable mixed terrain classification result with clear boundary information and will put the terrain depending gait selection and path planning of the multi-legged robots into practice.

Highlights

A multi-legged robot that originates from bionic of reptiles has high walking stability and low energy consumption in a stationary state
The SLIC segmentation technology is used to complete the we proposed a synthetic classification method to obtain both advantages of terrain segmentation process, and the improved terrain identification methods are combined with segmentation methods and classification methods
The SLIC is used to divide the mixed terrain and capture the terrains boundary; the image is subjected to image segmentation and the SVM terrain classifier based on the SURF method is used for terrain classification

Summary

Introduction

A multi-legged robot that originates from bionic of reptiles has high walking stability and low energy consumption in a stationary state. As a multi-legged robot represents a nonlinear, multi-body, rigid–flexible system having the complex interactions with the environment, the environmental characteristics have a great influence on robot mobility. Correct perception and ability to classify the terrain are necessary to make the correct gait planning, path planning and motion control strategy in time. To ensure robot adaptability to the environment and its ability to independently choose the region, and avoid the problems in stability movement control such as slipping and instability in the process of motion, it is necessary to improve robot ability to perceive different terrain characteristics. During the interaction between the robot and environment, both geometric and non-geometric features of the terrain influence the robot’s performance.

Methods

Findings

Discussion

Conclusion