‹ 연구 목록
건설현장 자동화를 위한 4D 시뮬레이션 및 강화학습 기반 타워크레인 자율 운행 기초 기술 개발
Autonomous Tower Crane Development based on 4D Simulation Environment and Reinforcement Learning for Construction Automation
- 연구 기간
- 2021.04 – 2022.12
- 발주 기관
- 국토교통과학기술진흥원
연구 개요 Overview
본 연구는 건설산업에서 노동집약적 패러다임을 전환하고 건설장비 대상 자율주행 연구 기초기술을 확보하기 위한 “건설현장 자동화를 위한 4D 시뮬레이션 및 강화학습 기반 타워크레인 자율운행 기초 기술 개발”연구로, 건설현장을 4D 시뮬레이션 환경으로 구축하고 타워크레인 자율화기능을 강화학습 기반으로 설계, 실험 및 검증하는 것을 최종 목표로 함.
This study aims to develop 4D simulation and reinforcement learning-based tower crane autonomous operation basic technology to shift the labor-intensive paradigm in the construction industry and secure basic technology for self-driving research for construction equipment.
연구 내용 Research Content
본 연구는 건설현장을 4D 시뮬레이션 환경으로 구축하고 타워크레인 자율화기능을 시뮬레이션 환경에서 설계, 강화학습 및 검증을 통해 노동집약적인 건설산업 생산성과 안전성 향상을 목표로 함. 타워크레인 같은 건설장비들은 실제 환경 데이터를 수집함에 있어 어려움을 가지고 있고, 이를 해결하기 위해 High-Fidelity Simulation을 개발함. 작업계획 및 자재 등을 연계한 4D 시뮬레이션 환경을 통해 타워크레인의 동작요소를 반영한 운행 모델링, 모방학습 인터페이스 개발 및 데이터셋을 구축한 뒤 강화학습을 활용한 End-to-End 네트워크 타워크레인 자율화 작업기능을 개발하는 연구를 수행함. 타워크레인 같은 건설장비들은 실제 환경 데이터를 수집함에 있어 어려움을 가지고 있고, 이를 해결하기 위해 High-Fidelity Simulation을 개발함. 작업계획 및 자재 등을 연계한 4D 시뮬레이션 환경을 통해 타워크레인의 동작요소를 반영한 운행 모델링, 모방학습 인터페이스 개발 및 데이터셋을 구축한 뒤 강화학습을 활용한 End-to-End 네트워크 타워크레인 자율화 작업기능을 개발하는 연구를 수행함. 본 연구의 핵심 성과물은 A) 건설현장 4D 시뮬레이션 기반 타워크레인 운행환경 구축 기술, B) 강화학습을 활용한 타워크레인 자율화 작업기능 개발 기술, C) 타워크레인 자율운행 Virtual-Model, 세 가지로 도출됨. 본 연구의 결과물로서 개발된 타워크레인 자율화 기술 및 구축된 4D 시뮬레이션 환경을 활용하여 추후 건설현장 전반에서의 강화 및 기계학습을 위한 데이터 생성에 사용될 수 있으며, 타워크레인 이외에도 타 건설장비의 자율화 구현을 위한 기반이 되는 결과물을 도출 할 수 있음. 특히, 실제 건설현장 및 장비의 물리적 특성을 반영하는 프 레임워크를 제시하는 것으로 기존의 연구에서 가정치로 두었던 많은 변수들을 실제의 현장 상황에 맞게 설정할 수 있게 하여 다른 건설현장 시뮬레이션의 기초 기술로 활용될 수 있을 것으로 기대함. 또한, 건설현장 4D 시뮬레이션 기반 타워크레인의 자율화 기능의 프레임워크를 제시하여 건설산업 생산성 증진을 위한 건설장비 자동화 구현에 문제되는 각종 문제 및 이슈들을 해결할 수 있을 것으로 기대되며 이를 통해 건설산업의 생산성을 향상을 도모할 수 있음.
This study aims to build a construction site in a 4D simulation environment and improve labor-intensive construction industry productivity and safety through design, reinforcement learning, and verification of tower crane autonomy functions in a simulation environment.
Construction equipment such as tower cranes have difficulty collecting actual environmental data, and to solve this problem, High-Fidelity Simulation was developed. Through a 4D simulation environment that links work plans and materials, we conduct research to develop end-to-end network tower crane autonomous work functions using reinforcement learning after developing operation modeling, imitation learning interfaces, and datasets. Construction equipment such as tower cranes have difficulty collecting actual environmental data, and to solve this problem, High-Fidelity Simulation was developed. Through a 4D simulation environment that links work plans and materials, we conduct research to develop end-to-end network tower crane autonomous work functions using reinforcement learning after developing operation modeling, imitation learning interfaces, and datasets.
The core achievements of this study were derived in three ways:
A) construction site 4D simulation-based tower crane operation environment construction technology
B) tower crane autonomous work function development technology using reinforcement learning
C) tower crane autonomous operation Virtual-Model.
Using the developed tower crane autonomous technology and the built 4D simulation environment as the result of this study, it can be used to generate data for reinforcement and machine learning throughout the construction site in the future, and the foundation for the autonomous implementation of other construction equipment can be derived.
In particular, by presenting a framework that reflects the physical characteristics of actual construction sites and equipment, it is expected that many variables assumed in previous studies can be set to suit the actual field situation and be used as basic technology for simulation of other construction sites. In addition, it is expected to solve various problems and issues in automating construction equipment to improve productivity of the construction industry by presenting a framework of the autonomous function of the construction site 4D simulation-based tower crane.
SPONSOR
Korea Agency for Infrastructure Technology Advancement (Award Number : 21CTAP-C163785-01)
Construction equipment such as tower cranes have difficulty collecting actual environmental data, and to solve this problem, High-Fidelity Simulation was developed. Through a 4D simulation environment that links work plans and materials, we conduct research to develop end-to-end network tower crane autonomous work functions using reinforcement learning after developing operation modeling, imitation learning interfaces, and datasets. Construction equipment such as tower cranes have difficulty collecting actual environmental data, and to solve this problem, High-Fidelity Simulation was developed. Through a 4D simulation environment that links work plans and materials, we conduct research to develop end-to-end network tower crane autonomous work functions using reinforcement learning after developing operation modeling, imitation learning interfaces, and datasets.
The core achievements of this study were derived in three ways:
A) construction site 4D simulation-based tower crane operation environment construction technology
B) tower crane autonomous work function development technology using reinforcement learning
C) tower crane autonomous operation Virtual-Model.
Using the developed tower crane autonomous technology and the built 4D simulation environment as the result of this study, it can be used to generate data for reinforcement and machine learning throughout the construction site in the future, and the foundation for the autonomous implementation of other construction equipment can be derived.
In particular, by presenting a framework that reflects the physical characteristics of actual construction sites and equipment, it is expected that many variables assumed in previous studies can be set to suit the actual field situation and be used as basic technology for simulation of other construction sites. In addition, it is expected to solve various problems and issues in automating construction equipment to improve productivity of the construction industry by presenting a framework of the autonomous function of the construction site 4D simulation-based tower crane.
SPONSOR
Korea Agency for Infrastructure Technology Advancement (Award Number : 21CTAP-C163785-01)
