Publications

Detailed Information

Trajectory generation for autonomous excavators based on expert operator forceing pattern : 자동화 굴착기를 위한 숙련자 굴착력 패턴 기반 굴착 작업 궤적 생성

DC Field Value Language
dc.contributor.advisor이동준-
dc.contributor.author김창묵-
dc.date.accessioned2020-10-13T02:38:28Z-
dc.date.available2020-10-13T02:38:28Z-
dc.date.issued2020-
dc.identifier.other000000162197-
dc.identifier.urihttps://hdl.handle.net/10371/169129-
dc.identifier.urihttp://dcollection.snu.ac.kr/common/orgView/000000162197ko_KR
dc.description학위논문 (석사) -- 서울대학교 대학원 : 공과대학 기계공학부, 2020. 8. 이동준.-
dc.description.abstractIn this thesis, we propose an excavation trajectory generation framework for
autonomous excavators based on expert operator forcing pattern. The primary
focus is to develop autonomous excavator system which is stable and guarantees
a certain quantity of excavation in various surroundings. We nd the excavation
trajectories based on the terrain features and the excavation forcing patterns
from the excavation data of expert operators. The expert excavation trajectories
are encoded with dynamic movement primitives (DMP) and learn through multilayer
perceptron (MLP). The excavation trajectory is generated according to the
terrain feature using the trained model. The excavator is modeled with 3-DoF
rigid body system, and the excavation force on the bucket tip is estimated online
by using the momentum-based disturbance observer(DOB). The estimated force
is added to the DMP as a coupling term to modulate the excavation trajectory
in real-time so that the estimated force can follow the expert excavation force
pattern. Lastly, we verify the performance of the suggested framework through
simulation and actual excavator test.
-
dc.description.abstract본 논문에서는 자동화 굴착기를 위한 숙련자 굴착력 패턴 기반 굴착 작업 궤적 계획 프레임워크를 제시한다. 본 프레임워크는 자동화 굴착기의 다양한 작업 환경에서 숙련자와 유사하게 안정된 굴착 작업을 수행하며, 굴착량이 보장되는 작업을 하는 것이 목표이다. 우선 숙련된 굴착기 작업자들의 굴착 작업 데이터로부터 지형 특징에 기반한 작업 궤적과 굴착력 패턴을 찾아내었다. 숙련자의 굴착 궤적은 dynamic movement primitives(DMP)으로 encoding하여 neural network의 한 기법인 multi-layer perceptron(MLP)을 통해 학습하고, 학습된 모델을 기반으로 지형에 따른 굴착 궤적을 생성하였다. 굴착기를 다자유도 강체 시스템으로 모델링 하고, 실시간으로 버켓 끝단에 걸리는 굴착력을 momentum-based disturbance observer를 이용하여 추정하였다. 추정된 굴착력은 실시간으로 굴착 궤적을 재생성 하기위해 DMP에 coupling term으로 추가하였고, 이를 통해 추정되는 굴착력이 숙련자의 굴착 패턴을 따라갈 수 있도록 제어하였다. 마지막으로 제안한 프레임워크에 대해서는 시뮬레이션 실험과 실제 굴착기를 이용한 실험을 통해 정합성 검증을 진행하였다.-
dc.description.tableofcontents1 Introduction 1
1.1 Motivation and Objectives 1
1.2 Related Work 2
1.3 Contribution 4
2 Preliminary 6
2.1 System Description 6
2.2 Excavator Dynamic Modeling 7
2.3 Force Estimation via Momentum Based Disturbance Observer 9
2.4 Dynamic Movement Primitives 10
3 Excavation Trajectory Generation 13
3.1 Analysis of Expert's Excavation Trajectory 13
3.2 Generate Nominal Excavation Trajectory by Imitating Expert Operator 19
3.3 Modulate Excavation Trajectory by Force Pattern of Expert Operator 22
4 Experiments 26
4.1 Excavation Simulation 26
4.1.1 Excavation on Flat and Slope Terrain 26
4.1.2 Excavation using Trajectory Generated by Incorrect Terrain Recognition 31
4.1.3 Excavation with Obstacle in the Ground 33
4.2 Excavation Test Result using Excavator 35
5 Conclusion and Future Work 40
5.1 Conclusion 40
5.2 Future Work 41
-
dc.language.isoeng-
dc.publisher서울대학교 대학원-
dc.subjectAutonomous excavators-
dc.subjectTrajectory generation-
dc.subjectDynamic movement primitives-
dc.subjectMulti-layer perceptron-
dc.subjectDynamics-
dc.subjectMomentum-based observer.-
dc.subject.ddc621-
dc.titleTrajectory generation for autonomous excavators based on expert operator forceing pattern-
dc.title.alternative자동화 굴착기를 위한 숙련자 굴착력 패턴 기반 굴착 작업 궤적 생성-
dc.typeThesis-
dc.typeDissertation-
dc.contributor.AlternativeAuthorKim, Changmuk-
dc.contributor.department공과대학 기계공학부-
dc.description.degreeMaster-
dc.date.awarded2020-08-
dc.identifier.uciI804:11032-000000162197-
dc.identifier.holdings000000000043▲000000000048▲000000162197▲-
Appears in Collections:
Files in This Item:

Altmetrics

Item View & Download Count

  • mendeley

Items in S-Space are protected by copyright, with all rights reserved, unless otherwise indicated.

Share