Page - 140 - in Proceedings - OAGM & ARW Joint Workshop 2016 on "Computer Vision and Robotics“
Image of the Page - 140 -
Text of the Page - 140 -
robotic tasks. The domain contains information about all the objects which can appear in the envi-
ronment. Further it knows about the skills the robot is able to perform. A skill is defined through its
nameandthe itsparameters. Askillhasapreconditionandaneffect. Theprecondition is thestate the
environmentneeds tobe inbefore theactioncanbeperformed. Theeffectofa skill is thedescription
of theenvironmentafter the skill isperformed.
Theproblemdescribes the initial states0 and thegoalg,whichareasetofpropositions [4]. Thegoal
state sg is a state, that satisfies g. First object instances are defined, which occur in the environment
and their initial properties are stored. The planner takes the domain and the problem description and
generates a list of skills, which need to be performed to solve the given problem. We use the planner
SGPlan6 [5]. This list of skills is forwarded to theexecutive layerof the3-TIERarchitecture.
For solving the order picking task, the following skills are required: moveBoxFromLevelToTray,
movBoxFromTrayToLevel and graspItem. Lets assume an environment containing transport boxes
BOXA,BOXC and a shelf with levelsLEVEL 1,LEVEL 2 and the goal of picking one item from
BOXCplacing it at thedeliveryboxDBOXCandpicking twoitemsfromBOXAplacing it atdeliv-
eryboxDBOXA.Theplannercomesupwith the followingplan(seeListing1).Thenameof theskill
is the first parameter, followed by the parameters the skill requires. So the first skill, which has to be
performed ismoveBoxFromLevelToTray. TheBOXC ismoved fromLEVEL 2 to theTRAY.
Listing 1: Outputof planner for exampledomainandproblem.
0 (MOVEBOXFROMLEVELTOTRAY BOX C LEVEL 2 TRAY)
1 (GRASPITEM BOX C DBOX C TRAY)
2 (MOVEBOXFROMTRAYTOLEVEL BOX C LEVEL 2 TRAY)
3 (MOVEBOXFROMLEVELTOTRAY BOX A LEVEL 1 TRAY)
4 (GRASPITEM BOX A DBOX A TRAY)
5 (GRASPITEM BOX A DBOX A TRAY)
6 (MOVEBOXFROMTRAYTOLEVEL BOX A LEVEL 1 TRAY)
4.2. ExecutiveLayer
The executive layer receives a list of skills from the planner. The executive layer handles the execu-
tion of single skills. Each skill is composed of skill primitives, which are the fundamental building
blocks of each skill. The executive layer knows about this decomposition and ensures that primi-
tives are executed in right order to guarantee a successful skill execution. This decomposition of the
skillsmoveBoxFromLevelToTray,moveBoxFromTrayToLevelandgraspItem is showninTable1. The
composition of skill primitives for each skill is intrinsic knowledge of this layer. Further it monitors
the outcome of each primitive. This layer has also the opportunity to perform recovery behaviors, if
primitives fail. If no recovery can be performed or the recovery fails, this failure is reported to the
planning layer. The decomposition of themoveBoxFromLevelToTray and its execution is shown in
Figure3.
skills moveBoxToRack graspItem moveBoxToLevel
skill
primi-
tives detectHandle detectItem detectHandle
graspHandle graspItem graspHandle
moveArmToSupportPose deliverItem moveArmToSupportPose
pullBox pullBox
moveBox moveBox
deliverBoxOnTray deliverBoxOnLevel
Table1: Within this table the skill primitivecompositionofall skills are listed.
4
140
Proceedings
OAGM & ARW Joint Workshop 2016 on "Computer Vision and Robotics“
- Title
- Proceedings
- Subtitle
- OAGM & ARW Joint Workshop 2016 on "Computer Vision and Robotics“
- Authors
- Peter M. Roth
- Kurt Niel
- Publisher
- Verlag der Technischen Universität Graz
- Location
- Wels
- Date
- 2017
- Language
- English
- License
- CC BY 4.0
- ISBN
- 978-3-85125-527-0
- Size
- 21.0 x 29.7 cm
- Pages
- 248
- Keywords
- Tagungsband
- Categories
- International
- Tagungsbände