Slide 1

Natchanon Wongwilai Adviser: Nattee Niparnan, Ph.D. M.Eng. 1

Slide 2

Introduction How to grasp?, Why failed to grasp?, Goal Related Works Vision-based grasping, Manipulation under uncertainty Our Problem Challenge, Proposed method Etc. Scopes, Work plan 2

Slide 3

[http://spectrum.ieee.org/robotics/robotics-software/slideshow-born-bionic/0] !? 3 ? ? ? ? ? ? Model = 0.53 W = 39 g

Slide 4

4

Slide 5

2D [Borst el at.,00; Chinellato el at.,05; Calli el at.,11;...] 3D [Miller el at.,03; Goldfeder el at.,07; Hubner el at.,08;...] 2.5D [Richtsfeld el at.,08;...] Others [Saxena el at.,08;...] 5

Slide 6

6 (Video)

Slide 7

The most common failure mode I've seen is that the closing fingers bump the object so that the fingers don't touch the intended contact points. Then the fingers knock the object completely out of the grasp. I think the causes are localization errors from the perception system and asking the robot to carry out an inherently dynamic task that was planned with static analysis tools. Jeff Trinkle GRSSP Workshop 2010 7 The most common failure mode I've seen is that the closing fingers bump the object so that the fingers don't touch the intended contact points. Then the fingers knock the object completely out of the grasp. I think the causes are localization errors from the perception system and asking the robot to carry out an inherently dynamic task that was planned with static analysis tools. Jeff Trinkle GRSSP Workshop 2010

Slide 8

Contact position error Theory vs. Practical Cause of error Sensor Control Computation Uncertainty 8 [http://www.cs.columbia.edu/~cmatei/]

Slide 9

Accuracy of fingertip placement Planning Using camera 9

Slide 10

SensorPriceAccuracyData type Tactile sensorExpensiveHighForce array Laser range finderExpensiveHighRange CameraVaryModerateImage Tactile sensor [Bekiroglu el at., 11] Laser range finderCamera 10

Slide 11

Vision-based grasping Stereo camera Eye-in-hand camera Manipulation under uncertainty Independent contact region Visual servoing Reactive grasping Probabilistic model 11

Slide 12

Stereo vision based grasping [Popovic et al.,11; Gratal el at., 12] 12

Slide 13

Eye-in-hand camera [Walck el at., 10; Lippiello el at., 11; Calli el at., 11] 13

Slide 14

14 (Video)

Slide 15

Independent contact region [Nilwatchararang et al., 08; Roa et al.,09] 15

Slide 16

Visual servoing [Gratal el at., 12; Calli el at., 11] 16

Slide 17

Reactive Grasping [Teichmann et al.,94; Hsiao et al.,09; Hsiao et al.,10] 17

Slide 18

Probabilistic model [Laaksonen et al.,11; Dogar et al.,11; Platt et al.,11] 18

Slide 19

Propose online planning method for accurate fingertip placement under uncertainty using eye-in-hand camera 19

Slide 20

ACCURACY!!! Insufficient information Bearing-only data Unknown object model and properties Dont have any initial information Close-up view with featureless image Kinematic constraint Unreachable position Object out of view Uncertainty Unpredictable noise 20

Slide 21

21

Slide 22

ModelingGrasp planningLocalizationGrasping 22

Slide 23

Grasping Localization Modeling Grasp planning 23

Slide 24

Robot build up a map and localize itself simultaneously while traversing in an unknown environment [Paul Newman, 06] 24

Slide 25

Robot locationHand(Fingertips) location Environment mapObject model 25

Slide 26

http://www.biorobotics.org/projects/tslam/experiments/slam1experiment.html 26

Slide 27

Probabilistic SLAM [Smith and Cheeseman, 86] The probability distribution of robot state and landmark locations The observation model The motion model 27

Slide 28

SLAM recursive algorithm Time-update Measurement Update 28

Slide 29

Feature detection Point features, Line features Feature association How features associate with landmarks Feature measurements Observation model 29 [http://www.sciencedirect.com/science/article/pii/S0377042711002834]

Slide 30

How to represent a map (object model) from available features 30 [http://www.deskeng.com/articles/aaayex.htm]

Slide 31

Exploration How to explore for object modeling Strategy Close-up strategy Out of view strategy 31

Slide 32

Fingertips placement evaluation Using ground truth data Contact position marking Modeling evaluation Using ground truth data from structural environment Database Kinect 32

Slide 33

Develop online planning method for accurate fingertip placement using eye-in-hand camera Not develop algorithm to find grasping points No clutter in work space Simple & Textured object 33

Slide 34

Study the works in the related fields Develop algorithms Test the system Evaluate a result Prepare and engage in a thesis defense 34

Slide 35

35