Pedro mitsuo shiroma advisors mario f m campos vijay kumar
Download
1 / 70

CoMutaR : An approach for task allocation - PowerPoint PPT Presentation


  • 63 Views
  • Uploaded on

Pedro Mitsuo Shiroma Advisors : Mario F. M. Campos Vijay Kumar. CoMutaR : An approach for task allocation. Motivation. In a near future it is very likely that several robots will be present in the same workspace. Motivation. In this scenario we will have:

loader
I am the owner, or an agent authorized to act on behalf of the owner, of the copyrighted work described.
capcha
Download Presentation

PowerPoint Slideshow about ' CoMutaR : An approach for task allocation' - martin-stein


An Image/Link below is provided (as is) to download presentation

Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.


- - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - -
Presentation Transcript
Pedro mitsuo shiroma advisors mario f m campos vijay kumar

Pedro Mitsuo Shiroma

Advisors: Mario F. M. Campos

Vijay Kumar

CoMutaR: An approach for task allocation


Motivation
Motivation

In a near future it is very likely that several robots will be present in the same workspace


Motivation1
Motivation

In this scenario we will have:

  • Robots executing its assigned tasks,

    • with rich sensorial capacity,

    • high processing power,

    • and communication devices;

  • A lot of underutilized resources;

  • Increasingly large amounts of continuously gathered data;


Motivation2
Motivation

How to effectively allocate the available resources (whole robots, sensors, gathered data, actuators, processing power, etc.) for the successful execution of multiple tasks?

or, more formally…


Problem definition
Problem definition

Given a set of (possibly) heterogeneous mobile robots located in a dynamical environment and a set of tasks that need to be executed, allocate tasks to one or more robots as efficiently as possible.


Problem definition1

robot1

Task1

robot2

Task2

Task3

robot3

Task4

...

...

robotm

Taskp

Problem definition


Outline
Outline

  • Related work

  • CoMutaR

  • Experimental results

  • Conclusion


Outline1
Outline

  • Related work

  • CoMutaR

  • Experimental results

  • Conclusion


Related work taxonomy
Related Work: Taxonomy

  • Tasks

  • Robots

  • Assignment

    [Gerkey 2003]


Related work taxonomy tasks
Related Work: Taxonomy – Tasks

  • SR – Single-Robot tasks

  • MR – Multi-Robot tasks

robot1

Task1

robot2

Task2

robot3


Related work taxonomy assignment
Related Work: Taxonomy – Assignment

  • No knowledge

  • Model-based

Task1

Task2

t

t0

t1

Task1

t2

Task2

t


Related work taxonomy robots
Related Work: Taxonomy – Robots

  • ST – Single-Task Robot

  • MT – Multi-Task Robot

robot1

Task1

robot2

Task2

Task3



Related work1
Related work

“'Most commonly, task allocation problems assume robots are single-task robots, since more capable robots that perform multiple tasks in parallel are still beyond the current state of the art”

Lynne E. Parker, Springer handbook of robotics (2008)

“The MT-SR-IA and MT-SR-TA problems are currently uncommon, as they assume robots that can each concurrently execute multiple tasks. Today’s mobile robots are generally actuator-poor. Their ability to affect the environment is typically limited to changing position, so they can rarely execute more than one task at a time.”

Gerkey & Mataric – Int. J. of Robotics Research (2004)


Outline2
Outline

  • Related work

  • CoMutaR

  • Experimental results

  • Conclusion


Comutar co alition formation based on mu lti ta sking r obots
CoMutaR: Coalition formation based on Multitasking Robots

  • MT-MR-IA;

  • Coalition of actions x schemas

  • Queries x information type

  • Share-restricted resources x



Comutar overview1
CoMutaR: Overview

  • Assumptions

    • Tasks can be subdivided into actions;

    • Actions are able to determine or estimate, based on the query, its constraint functions;

    • If all inputs (queries) and constraint functions are not broken then the action will successfully perform as expected.


Comutar nomenclature
CoMutaR: Nomenclature

robot

  • : the 4th action in robot 3;

  • : the 2nd share-restricted resource in robot 1

  • : the constraint function imposed by

    over

action

Share-restricted resource


Comutar
CoMutaR

  • Overview

  • Action

  • Coalition

  • Query

  • Share-restricted resource


Comutar action

action2

robot1

action1

action3

CoMutaR: Action

  • An action is any computational module that can either produce data or perform a task

robot1

Task1


Comutar action1
CoMutaR: Action

  • Sensor reading:

  • Data transform:

  • Task accomplishing:

actioni

actioni

actioni

Task


Comutar1
CoMutaR

  • Overview

  • Action

  • Coalition

  • Query

  • Share-restricted resource


Comutar coalition
CoMutaR: Coalition

  • Coalition is a temporary organization of actions that are brought together to tackle a particular task

obstacle

detector

laser

obstacle

avoider

Obstacle

avoidance

obs

range


Comutar bidding protocol

open auction

open query

open query

cost

cost

bid

winner

connect

CoMutaR: Bidding protocol

  • Coalition formation

  • Protocol (IA)

Obstacle

avoidance

obstacle avoider

obstacle detector

laser

obs

range

sonar

range

Obstacle avoidance

obstacle avoider

obstacle detector

laser

sonar

t

connect


Comutar2
CoMutaR

  • Overview

  • Action

  • Coalition

  • Query

  • Share-restricted resource


Comutar query
CoMutaR: Query

  • Information type [Donald1997]:

  • Resource database [Cowley2004]:

  • Query:

obstacle

detector

laser

range


Comutar query1
CoMutaR: Query

  • Dynamical environments:

  • Reactive tasks:

  • Mapping:

  • User interface:


Comutar query2

data importer

obstacle detector

Laser

data exporter

obs

range

robotj

CoMutaR: Query

obstacle avoider

obstacle detector

Laser

Obstacle

avoidance

range

obs

roboti


Comutar query3
CoMutaR: Query

  • An action can execute only if all input queries are answerable;

  • A query is a contract between producer and consumer, which guarantee data connection;

  • A query can include other queries:

    • A surveillance action is taking snapshots:

    • A pursue action in the same robot is trying to track an intruder:


Comutar query4
CoMutaR: Query

  • An action can intervene on the query of another action:

    • An obstacle avoidance is querying for obstacles:

    • A box pushing action can alter the region queried by the obstacle avoidance:


Comutar3
CoMutaR

  • Overview

  • Action

  • Coalition

  • Query

  • Share-restricted resource


Comutar share restricted resource
CoMutaR: Share-restricted resource

  • MT (Multitasking) robots

Task1

robot1

Task2


Comutar share restricted resource1
CoMutaR: Share-restricted resource

  • A share-restricted resource is any property that belongs to a robot and cannot be freely shared among actions.

    • Communication link;

    • Processing power;

    • Pose;

    • Power consumption;

  • A share-restricted resource is any property that belongs to a robot and cannot be freely shared among actions.

    • Communication link;

    • Processing power;

    • Pose;

    • Power consumption;


Comutar share restricted resource pose
CoMutaR: Share-restricted resource – Pose

  • Configuration space

composition

surveillance

controller


Comutar share restricted resource pose1
CoMutaR: Share-restricted resource – Pose

  • Constraints imposed by sensors

compound

Planned task for

another action

Region to be sensored


Comutar share restricted resource constraint function

action1,2

...

CoMutaR: Share-restricted resource – Constraint function

...

action1,1

...

...

...

. . .

...

actioni,n

...


Comutar share restricted resource constraint function1
CoMutaR: Share-restricted resource – Constraint function

  • Codomain;

  • Constraint function:

  • Maximum capacity:

  • compound operator:

  • comparison operator:


Comutar share restricted resource communication link
CoMutaR: Share-restricted resource – Communication link

  • codomain:

  • : required communication bandwidth

  • : maximum bandwidth

  • 30Mb/s + 20Mb/s ≤ 100Mb/s


Comutar share restricted resource processing power
CoMutaR: Share-restricted resource – Processing power

  • codomain:

  • : required processing time

  • : allocated processing power

  • 10Mhz + 20Mhz ≤ 100Mhz


Comutar share restricted resource pose2
CoMutaR: Share-restricted resource – Pose

  • Configuration space

Obstacle avoider

Non-holonomic constraint

composition


Comutar share restricted resource pose3
CoMutaR: Share-restricted resource – Pose

  • Constraints imposed by sensors (queries)

Queries are periodically updated


Comutar share restricted resource pose4

Potential field

CoMutaR: Share-restricted resource – Pose

  • Scenario



Comutar co alition formation based on mu lti ta sking r obots1
CoMutaR: Coalition formation based on Multitasking Robots

action2

action1

Task1

db1

db2

actionn


Outline3
Outline

  • Related work

  • CoMutaR

  • Experimental results

  • Conclusion



Tightly coupled tasks box pushing1

Trajectory

Tightly-coupled tasks – Box pushing

  • Scenario




Comparison with st approaches
Comparison with ST approaches

MT – Transportation and surveillance with task release

ST – Transportation followed by surveillance

MT – Transportation and surveillance





Outline4
Outline

  • Related work

  • CoMutaR

  • Experimental results

  • Conclusion


Contributions
Contributions

  • Multitasking robots;

  • Query + subdivision in actions:

    • task feasibility

  • Definition of share-restricted resources and constraint functions;

  • Related Publications:

    • Shiroma, P. M., Campos, M. F. M.; CoMutaR: A framework for multi-robot coordination and task allocation; IROS 2009;

    • Shiroma, P. M., Campos, M. F. M.; Multi-robot task allocation through share-restricted resources; SBAI 2009;

    • Garcia, R. F., Shiroma, P. M., Chaimowicz, L., Campos, M. F. M.; Um arcabouço para localização de enxame de robôs; SBAI 2007


Conclusion
Conclusion

  • CoMutaR was successfully tested in 4 tasks;

  • Multitasking is feasible;

  • Guarantee of data connection;

  • MT-MR-IA;

  • A robot is able to continue executing a task even in the presence of faulty sensors;

  • Subdivision in actions allow us to easily define the constraint functions and consequently check for conflicts;


Future works
Future works

  • Define a winner policy;

  • Experiments with more tasks;

  • Include planning (optimization techniques);



Related work main approaches
Related Work: Main approaches

  • Market-based

    • Task announcement;

    • Winner determination.



Related work2
Related Work:

  • MURDOCH

    • Market-based;

    • Resources;

    • ST-SR-IA;

  • ASyMTRe

    • Automated coalition formation;

    • Information type;

    • Does not guarantee link persistence;

  • RACHNA

    • Coalition-based



Preliminary results box pushing localization system
Preliminary results – Box pushing Localization system

  • ARToolkitPlus [Wagner 2007];

  • 3 cameras;

  • Up to 40 markers;

  • 3D pose.


Preliminary results box pushing localization system1
Preliminary results – Box pushingLocalization system

  • Robust localization;

  • UKF [Vilar 2007].


Preliminary results box pushing

Overhead camera

Lateral view

Preliminary results – Box pushing





Resultados experimentais transporte
Resultados Experimentais – Transporte

trajetória

Setup experimental

Espaço de conf. projetado


ad