Abstraction as a Tool to Bridge the Reality Gap in Evolutionary Robotics

Kirk Scheper

doi:10.4233/uuid:389f453e-f7ff-4fea-a353-32755cf9a9e1

Abstraction as a Tool to Bridge the Reality Gap in Evolutionary Robotics

Kirk Scheper

Control & Simulation

Research output: Thesis › Dissertation (TU Delft)

104 Downloads (Pure)

Abstract

Automatically optimizing robotic behavior to solve complex tasks has been one of
the main, long-standing goals of Evolutionary Robotics (ER). When successful, this
approach will likely fundamentally change the rate of development and deployment
of robots in everyday life. Performing this optimization on real robots can be risky
and time consuming. As a result, much of the work in ER is done using simulations
which can operate many times faster than realtime. The only downside of this, is
that, due to the limited fidelity of the simulated environment, the optimized robotic
behavior is typically different when transferred to a robot in the real world. This
difference is referred to as the reality gap...

Original language	English
Qualification	Doctor of Philosophy
Awarding Institution	Delft University of Technology
Supervisors/Advisors	de Croon, G.C.H.E., Supervisor Mulder, M., Supervisor
Award date	10 Sept 2019
Electronic ISBNs	978-94-6366-197-3
DOIs	https://doi.org/10.4233/uuid:389f453e-f7ff-4fea-a353-32755cf9a9e1
Publication status	Published - 10 Sept 2019

Keywords

Evolutionary Robotics
Reality Gap
Abstraction
Robust behavior
MAV

Access to Document

10.4233/uuid:389f453e-f7ff-4fea-a353-32755cf9a9e1

kywscheper_dissertationFinal published version, 36.6 MB

2 Article
1 Conference contribution
1 Special issue

Abstraction, Sensory-Motor Coordination, and the Reality Gap in Evolutionary Robotics
Scheper, K. & de Croon, G., 2017, In: Artificial Life. 23, 2, p. 124-141 18 p.
Research output: Contribution to journal › Article › Scientific › peer-review

Open Access
File
7 Citations (Scopus)

37 Downloads (Pure)
Vertical Landing for Micro Air Vehicles using Event-Based Optical Flow
Scheper, K., de Croon, G. & Pijnacker Hordijk, B. J., 15 Dec 2017, In: Journal of Field Robotics. 35, 1, p. 69-90
Research output: Contribution to journal › Special issue › Scientific › peer-review
41 Citations (Scopus)
Abstraction as a Mechanism to Cross the Reality Gap in Evolutionary Robotics
Scheper, K. & de Croon, G., 10 Aug 2016, From Animals to Animats 14. Tuci, E., Giagkos, A., Wilson, M. & Hallam, J. (eds.). Springer, Vol. 9825. p. 280-292 12 p. (Lecture Notes in Computer Science; vol. 9825).
Research output: Chapter in Book/Conference proceedings/Edited volume › Conference contribution › Scientific › peer-review

Open Access
File
6 Citations (Scopus)

80 Downloads (Pure)

Cite this

@phdthesis{389f453ef7ff4feaa35332755cf9a9e1,

title = "Abstraction as a Tool to Bridge the Reality Gap in Evolutionary Robotics",

abstract = "Automatically optimizing robotic behavior to solve complex tasks has been one ofthe main, long-standing goals of Evolutionary Robotics (ER). When successful, thisapproach will likely fundamentally change the rate of development and deploymentof robots in everyday life. Performing this optimization on real robots can be riskyand time consuming. As a result, much of the work in ER is done using simulationswhich can operate many times faster than realtime. The only downside of this, isthat, due to the limited fidelity of the simulated environment, the optimized roboticbehavior is typically different when transferred to a robot in the real world. Thisdifference is referred to as the reality gap...",

keywords = "Evolutionary Robotics, Reality Gap, Abstraction, Robust behavior, MAV",

author = "Kirk Scheper",

year = "2019",

month = sep,

day = "10",

doi = "10.4233/uuid:389f453e-f7ff-4fea-a353-32755cf9a9e1",

language = "English",

type = "Dissertation (TU Delft)",

school = "Delft University of Technology",

}

TY - THES

T1 - Abstraction as a Tool to Bridge the Reality Gap in Evolutionary Robotics

AU - Scheper, Kirk

PY - 2019/9/10

Y1 - 2019/9/10

N2 - Automatically optimizing robotic behavior to solve complex tasks has been one ofthe main, long-standing goals of Evolutionary Robotics (ER). When successful, thisapproach will likely fundamentally change the rate of development and deploymentof robots in everyday life. Performing this optimization on real robots can be riskyand time consuming. As a result, much of the work in ER is done using simulationswhich can operate many times faster than realtime. The only downside of this, isthat, due to the limited fidelity of the simulated environment, the optimized roboticbehavior is typically different when transferred to a robot in the real world. Thisdifference is referred to as the reality gap...

AB - Automatically optimizing robotic behavior to solve complex tasks has been one ofthe main, long-standing goals of Evolutionary Robotics (ER). When successful, thisapproach will likely fundamentally change the rate of development and deploymentof robots in everyday life. Performing this optimization on real robots can be riskyand time consuming. As a result, much of the work in ER is done using simulationswhich can operate many times faster than realtime. The only downside of this, isthat, due to the limited fidelity of the simulated environment, the optimized roboticbehavior is typically different when transferred to a robot in the real world. Thisdifference is referred to as the reality gap...

KW - Evolutionary Robotics

KW - Reality Gap

KW - Abstraction

KW - Robust behavior

KW - MAV

U2 - 10.4233/uuid:389f453e-f7ff-4fea-a353-32755cf9a9e1

DO - 10.4233/uuid:389f453e-f7ff-4fea-a353-32755cf9a9e1

M3 - Dissertation (TU Delft)

ER -

Abstraction as a Tool to Bridge the Reality Gap in Evolutionary Robotics

Abstract

Keywords

Access to Document

Fingerprint

Research output

Abstraction, Sensory-Motor Coordination, and the Reality Gap in Evolutionary Robotics

Vertical Landing for Micro Air Vehicles using Event-Based Optical Flow

Abstraction as a Mechanism to Cross the Reality Gap in Evolutionary Robotics

Cite this