Beyond Geometric Path Planning: When Context Matters Outline Structured To Unstructured Environments

Beyond Geometric Path Planning: When Context Matters Outline Structured To Unstructured Environments www.phwiki.com

Beyond Geometric Path Planning: When Context Matters Outline Structured To Unstructured Environments

Sorrell, Charlie, Contributing Writer has reference to this Academic Journal, PHwiki organized this Journal Beyond Geometric Path Planning: When Context MattersAshesh Jain, Shikhar Sharma Thorsten Joachims in addition to Ashutosh SaxenaOutlineMotivationApproachContext-based scoreFeedback mechanismLearning algorithmResultsJain, Sharma, Joachims, SaxenaStructured To Unstructured Environments[Images from Google]Kuka armKivaBeamBaxterPR2Robot NurseJain, Sharma, Joachims, Saxena

St Joseph School of Nursing RI www.phwiki.com

This Particular University is Related to this Particular Journal

Path PlanningHigh DoF manipulatorsContinuous high dimensional spaceObstaclesBASE7 DoF armJointLinkEnd-EffectorABGeometric criteria’sCollision freeShortest pathLeast timeMinimum energyKavraki et. al. PRMLaValle et. al. RRTRatliff et. al. CHOMPKaraman et. al. RRTSchulman et. al. TrajOptJain, Sharma, Joachims, SaxenaContext Rich EnvironmentJain, Sharma, Joachims, SaxenaContext Rich EnvironmentJain, Sharma, Joachims, Saxenahttp://www.youtube.com/watchv=uLktpkd7ojAVideo [14 sec to 18 sec]

What went wrongRobot not modeling the contextDoes not underst in addition to the preferencesJain, Sharma, Joachims, SaxenaDoes Existing Works Address ThisInverse Rein as long as cement Learning(Kober in addition to Peters 2011 , Abbeel et. al. 2010 , Ziebrat et. al. 2008 , Ratliff et. al. 2006)Context is not important, focuses on specific trajectoryModeling human navigation patterns Kitani et. al. ECCV 2012Optimal Demonstrations: Requires an expertAbbeel et. al.Ratliff et. al.Kober et. al.Jain, Sharma, Joachims, SaxenaOur GoalModel ContextGenerate multiple trajectories as long as a taskUser preferencesLearn from non-expert’sJain, Sharma, Joachims, Saxena

OutlineMotivationApproachContext-based scoreFeedback mechanismLearning algorithmResultsJain, Sharma, Joachims, SaxenaLearning SettingUserRobotOnline learning systemLearns from user feedbackSub-optimal feedbackGoal: Learn user preferencesJain, Sharma, Joachims, SaxenaOutlineMotivationApproachContext-based scoreFeedback mechanismLearning algorithmResultsJain, Sharma, Joachims, Saxena

Example of PreferencesMove a glass of waterUprightContextContorted ArmPreferences varies with users, tasks in addition to environmentsJain, Sharma, Joachims, SaxenaScore functionRobot configuration in addition to Environment InteractionsContextTrajectoryJain, Sharma, Joachims, SaxenaObject-object InteractionsConnecting waypoints to neighboring objectsTrajectory graphScore functionJain, Sharma, Joachims, SaxenaTrajectory graphObject attributes: {electronic, fragile, sharp, liquid, hot, }E.g. Laptop: {electronic, fragile} Knife: {sharp} Hermans et. al. ICRA w/s 2011Koppula et. al. NIPS 2011Distance featuresObject-object Interactions

Score functionObject-object InteractionsRobot configuration in addition to Environment InteractionsBadGoodJain, Sharma, Joachims, SaxenaFeaturesSpectrogram Object’s distance from horizontal in addition to vertical surfacesObject’s angle with vertical axisRobot’s wrist in addition to elbow configuration in cylindrical co-ordinateCakmak et. al. IROS 2011OutlineMotivationApproachContext-based scoreFeedback mechanismLearning algorithmResultsJain, Sharma, Joachims, SaxenaUser FeedbackIntuitive feedback mechanismsRe-rankInteractiveZero-GJain, Sharma, Joachims, Saxena

1. Re-rankRobot ranks trajectories in addition to user selects oneTop three trajectoriesUser feedbackUser observing top three trajectoriesJain, Sharma, Joachims, Saxena2. Zero-GUser corrects trajectory waypointsBad waypoint in redHolding wrist activates zero-G modeJain, Sharma, Joachims, SaxenaBad waypoint in redHolding wrist activates zero-G mode2. Zero-GUser corrects trajectory waypointsJain, Sharma, Joachims, Saxena

3. InteractiveNot all robots support zero-G feedbackJain, Sharma, Joachims, Saxena3. InteractiveNot all robots support zero-G feedbackJain, Sharma, Joachims, SaxenaOutlineMotivationApproachContext-based scoreFeedback mechanismLearning algorithmResultsJain, Sharma, Joachims, Saxena

Sorrell, Charlie Wired News Contributing Writer www.phwiki.com

Coactive LearningUserRobotGoal: Learn user preferencesShivaswamy & Joachims, ICML 2012Learn from sub-optimal feedbackJain, Sharma, Joachims, SaxenaTrajectory Preference PerceptronRegret boundShivaswamy & Joachims, ICML 2012Jain, Sharma, Joachims, SaxenaOutlineMotivationApproachContext-based scoreFeedback mechanismLearning algorithmResultsJain, Sharma, Joachims, Saxena

Experimental SetupTwo robots: Baxter in addition to PR235 tasks in household setting2100 expert labeled trajectories16 tasks in grocery store checkout settings1300 expert labeled trajectories14 objectsBowl, Knife, Laptop, Metal box, Fruits, Egg cartons etc.7 usersJain, Sharma, Joachims, SaxenaExperimental Setting 1Household environment on PR2PouringCleaning the tableSetting up table35 tasks Variation in objects in addition to environment Expert’s label on 2100 trajectories on a scale of 1 to 5Jain, Sharma, Joachims, SaxenaExperimental Setting 2Grocery store checkout on BaxterCereal boxEgg cartonKnife in human vicinity16 tasks Variations in objects in addition to their placementExpert’s label on 1300 trajectories on a scale of 1 to 5Jain, Sharma, Joachims, Saxena

Thank YouFor more details visit http://pr.cs.cornell.edu/coactive

Sorrell, Charlie Contributing Writer

Sorrell, Charlie is from United States and they belong to Wired News and they are from  San Francisco, United States got related to this Particular Journal. and Sorrell, Charlie deal with the subjects like Computers; Information Technology Industry

Journal Ratings by St Joseph School of Nursing

This Particular Journal got reviewed and rated by St Joseph School of Nursing and short form of this particular Institution is RI and gave this Journal an Excellent Rating.