Miniature-Scale Radio-Controlled Excavator Robot: Expedient to Automated3D and 2D Poses Labeling

Daeho Kim; Saffari Sheida

doi:10.55785/JCAR.1.3.14

All Issue

2022 Vol.1, Issue 3 Preview Page Next Page

Miniature-Scale Radio-Controlled Excavator Robot: Expedient to Automated3D and 2D Poses Labeling 미니어처 엑스커베이터 로봇을 활용한 3D/2D 포즈 레이블링 자동화

30 September 2022. pp. 14-17

PDF XML

Abstract

Innovation in construction robotics directly aligns with the priorities of the Korean construction industries and internationally, which has struggled for decades with stagnant productivity, poor safety, and a shortage of skilled young laborers. To facilitate construction robotics research and development, the authors present the use of miniature-scale radio-controlled (RC) equipment (e.g., excavator) as an expedient to automated data labeling for equipment images and construction robot’s artificial intelligence (AI) demonstration. We started with an off-the-shelf miniature-scale RC excavator that has the same mechanical parts, connections, and the standard structure with a real-scale one and modified its RC system with additional payloads (e.g., Arduino and Bluetooth module) such that we can directly commend it for any yet precise motions to it from a laptop. This paper introduces a potential approach to leverage this inexpensive system in excavator pose labeling for both 3D and 2D.

Keywords

Miniature-Scale Radio-Controlled Excavator

Computer Vision

Deep Neural Network

Automated Data Annotation

생산성, 안정성, 수익률 향상, 그리고 젊은 노동인구의 부족 현상에 대응하기 위해선 robotic automation and digitization으로 대변되는 Industry 4.0으로의 기술혁신이 필요하다. 본 연구에서는 건설로봇 연구개발(R&D)을 지원하기 위한 하나의 수단으로써, miniature-scale radio-controlled(RC) equipment의 사용을 제안하고자 한다. 시각적 인공지능(Visual AI)은 Industry 4.0의 핵심 기술이지만, 방대한 양의 학습데이터를 요구한다. 하지만, 실 현장에서 작업중인 중장비의 데이터를 수집 및 레이블링 하는 것은 현실적으로 불가능 하다. 본 연구는 시중의 미니어처 엑스커베이터의 RC 시스템을 커스터마이즈하여, 기존의 컨트롤러를 사용하지 않고, PC를 통해 보다 구체적인 명령을 내릴 수 있도록 변경하였다. 버켓모션에 대한 사전검증 결과, 150 번의 시도에서 약 1-2° 내외의 mean average deviation(MAD)(|예측값 - 측정값|)를 확인 하였다. 본 연구에서는 이 경제적인 시스템을 활용한 중장비의 3D/2D 포즈 레이블링 방법을 제안하고자 한다.

키워드

미니어처 RC 엑스커베이터

컴퓨터 비전

딥뉴럴네트워크

레이블링 자동화

References

Akinosho, T.D., Oyedele, L.O., Bilal, M., et al. (2020). Deep learning in the construction industry: A review of present status and future innovations. J Build Eng, 32: 101827. 10.1016/j.jobe.2020.101827

Delgado, J.M.D., and Oyedele, L. (2021). Deep learning with small datasets: using autoencoders to address limited datasets in construction management. Appl Soft Comput, 112: 107836. 10.1016/j.asoc.2021.107836

Global Infrastructure Outlook (GIO). (2020). "Forecasting infrastructure investment needs and gaps." <https://outlook.gihub.org/ > (Mar. 12, 2021).

International Labour Organization (ILO). (2019). "World employment and social outlook." <https://www.ilo.org/global/about-the-ilo/multimedia/maps-and-charts/WCMS_337082/lang--en/index.htm > (Mar. 12, 2021).

Mckinsey Global Institute (MGI). (2017). "Reinventing construction: A route to higher productivity."

United Nation (UN). (2019). <https://www.un.org/development/desa/en/news/population/world-population-prospects-2019.html > (Mar. 12, 2021)

Xiao, B., Kang, S.-C. (2021). Development of an Image Data Set of Construction Machines for Deep Learning Object Detection. J Comput Civ Eng, 35: 05020005. 10.1061/(ASCE)CP.1943-5487.0000945

Information

Publisher :Korean Society of Automation and Robotics in Construction
Publisher(Ko) :(사)한국건설자동화·로보틱스학회
Journal Title :Journal of Construction Automation and Robotics
Journal Title(Ko) :건설자동화·로보틱스 논문집
Volume : 1
No :3
Pages :14-17
Received Date : 2022-09-26
Revised Date : 2022-09-29
Accepted Date : 2022-09-30
DOI :https://doi.org/10.55785/JCAR.1.3.14

Journal of Construction Automation and Robotics ISSN:2800-0552(Print) 2951-116X(Online) 건설자동화·로보틱스 논문집

All Issue