B659: Principles of B659: Principles of Intelligent Robot Intelligent Robot

MotionMotionKris HauserKris Hauser

TA: Mark WilsonTA: Mark Wilson

An intelligent robot must be able to An intelligent robot must be able to coordinate its own motionscoordinate its own motions in order to in order to achieve certain goalsachieve certain goals

Coordinating motion Coordinating motion demands intelligencedemands intelligence

How do cognition, learning, and reflex How do cognition, learning, and reflex interact to produce intelligent behavior?interact to produce intelligent behavior?

How do we encode this in a robot?How do we encode this in a robot?

Two (or Three) Core Two (or Three) Core ProblemsProblems

Planning: choosing present behavior Planning: choosing present behavior to attain future goalsto attain future goals

Sensing: making meaningful Sensing: making meaningful interpretations from raw datainterpretations from raw data

Models are the underlying Models are the underlying “knowledge” that are used for “knowledge” that are used for drawing conclusions about past, drawing conclusions about past, present, and futurepresent, and future

Intelligent Robot Intelligent Robot ArchitectureArchitecture

Percept (raw sensor data)

Action

?

Intelligent Robot Intelligent Robot ArchitectureArchitecture

Percept (raw sensor data)

Action

Sensing

Planning

Intelligent Robot Intelligent Robot ArchitectureArchitecture

Percept (raw sensor data)

Action

Sensing algorithms

Models

Planning

State estimationMappingTrackingCalibrationObject recognition…

Intelligent Robot Intelligent Robot ArchitectureArchitecture

Percept (raw sensor data)

Action

Sensing algorithms

Models

Real-Time Control

Task Planning Motion Planning

Goal

Planning TopicsPlanning Topics Motion planningMotion planning Compute Compute high-level strategieshigh-level strategies, e.g.:, e.g.:

• Geometric pathsGeometric paths• Time-parameterized trajectoriesTime-parameterized trajectories• Sequence of sensor-based motion commandsSequence of sensor-based motion commands

To To achieve high-level goalsachieve high-level goals, e.g.:, e.g.:• Go from A to B while avoiding obstaclesGo from A to B while avoiding obstacles• Assemble product PAssemble product P• Build a map of environment EBuild a map of environment E• Find object OFind object O

Planning Topics (cont.)Planning Topics (cont.)

Feedback ControlFeedback Control Compute (or verify) a real-time Compute (or verify) a real-time

strategy for responding to deviations strategy for responding to deviations from the desired pathfrom the desired path

Sensing TopicsSensing Topics

State estimationState estimation Calibrating the parameters of static Calibrating the parameters of static

models using motionmodels using motion• Cameras and motion captureCameras and motion capture• System identificationSystem identification

Filtering and estimating dynamic modelsFiltering and estimating dynamic models• State estimation and sensor fusionState estimation and sensor fusion• Localization and mappingLocalization and mapping

Learning from demonstrationLearning from demonstration

Modeling TopicsModeling Topics

Rigid transformationsRigid transformations Kinematics and inverse kinematicsKinematics and inverse kinematics Dynamics of articulated structuresDynamics of articulated structures Cameras and laser rangefindersCameras and laser rangefinders

Relationship to AIRelationship to AI

““Sub-symbolic” Sub-symbolic” intelligenceintelligence

Continuous domainsContinuous domains Computationally Computationally

complex: basic complex: basic problems are in problems are in PSPACEPSPACE

Goals of the ClassGoals of the Class

Present Present formal and practical formal and practical algorithmicalgorithmic and and mathematical toolsmathematical tools for interpreting and synthesizing for interpreting and synthesizing motionmotion

Components of a general frameworkComponents of a general framework for designing and studying complex for designing and studying complex robotic and biological agentsrobotic and biological agents

This class does NOT cover…This class does NOT cover…

Lagrangian mechanicsLagrangian mechanics Rigid body simulationRigid body simulation Control theoryControl theory Numerical optimizationNumerical optimization Computer visionComputer vision

… … but makes use of knowledge from but makes use of knowledge from these subjectsthese subjects

Fundamental question of Fundamental question of motion planningmotion planning

Are the two given points connected by a path?Are the two given points connected by a path?

Forbidden region

Feasible space

Fundamental question of Fundamental question of motion planningmotion planning

Are the two given points connected by a path?Are the two given points connected by a path?

Forbidden region

Feasible space

e.g.:collision with obstaclelack of stabilitylost visibility with target…

Basic ProblemBasic Problem

Statement:Statement: Compute a Compute a collision-free pathcollision-free path for a rigid or for a rigid or articulated object among static obstaclesarticulated object among static obstacles

Inputs:Inputs:• Geometry of moving object and obstaclesGeometry of moving object and obstacles• Kinematics of moving object (degrees of Kinematics of moving object (degrees of

freedom)freedom)• Initial and goal Initial and goal configurationsconfigurations (placements) (placements)

Output:Output:Continuous sequence of collision-free robot Continuous sequence of collision-free robot

configurations configurations connecting the initial and goal connecting the initial and goal configurationsconfigurations

Why is this hard?Why is this hard?

Tool: Configuration SpaceTool: Configuration Space

Problems:• Geometric complexity• Space dimensionality

Some VariantsSome Variants Moving obstaclesMoving obstacles Multiple robotsMultiple robots Movable objectsMovable objects Assembly planningAssembly planning Goal is to acquire Goal is to acquire

information by sensinginformation by sensing• Model buildingModel building• Object finding/trackingObject finding/tracking• InspectionInspection

Nonholonomic Nonholonomic constraintsconstraints

Dynamic constraintsDynamic constraints Stability constraintsStability constraints

Optimal planningOptimal planning Uncertainty in model, Uncertainty in model,

control and sensingcontrol and sensing Exploiting task Exploiting task

mechanics (sensorless mechanics (sensorless motions, under-motions, under-actuated systems)actuated systems)

Physical models and Physical models and deformable objectsdeformable objects

Integration of planning Integration of planning and controland control

Integration with Integration with higher-level planninghigher-level planning

ApplicationsApplications

HRP-2, AIST, Japan

Humanoid RobotsHumanoid Robots

Lunar Vehicle (ATHLETE, Lunar Vehicle (ATHLETE, NASA/JPL)NASA/JPL)

Climbing RobotClimbing Robot

Design for Manufacturing Design for Manufacturing and Servicingand Servicing

General Electric

General MotorsGeneral Motors

Manipulation of Deformable Manipulation of Deformable ObjectsObjects

Manipulation of Deformable Manipulation of Deformable ObjectsObjects

Assembly SequencingAssembly Sequencing

Assembly SeqencingAssembly Seqencing

Cable Harness/Pipe DesignCable Harness/Pipe Design

Where to move next?

Map BuildingMap Building

Navigation Through Virtual Navigation Through Virtual EnvironmentsEnvironments

Computer-Assisted Computer-Assisted Angiography/ Colonoscopy/ Angiography/ Colonoscopy/

BronchoscopyBronchoscopy

CyberKnife (Accuray)

Radiosurgical PlanningRadiosurgical Planning

Self-ParkingSelf-Parking

Kineo

Transportation of A380 Transportation of A380 Fuselage through Small Fuselage through Small

VillagesVillages

Inhibitor binding to HIV protease

Study of Motion of Bio-Study of Motion of Bio-MoleculesMolecules

PrerequisitesPrerequisites

Ability and willingness to complete a Ability and willingness to complete a significant programming project on a significant programming project on a simulation GUI or physical robotsimulation GUI or physical robot

Interest in reading and discussing research Interest in reading and discussing research papers each weekpapers each week

Subjects: linear algebra*, multivariable Subjects: linear algebra*, multivariable calculus, geometry, probabilitycalculus, geometry, probability

Basic knowledge and taste for geometry, Basic knowledge and taste for geometry, mathematical analysis, and algorithmsmathematical analysis, and algorithms

BookBook

Principles of Robot Motion Principles of Robot Motion (Choset, (Choset, Hutchinson, Kantor, Burgard, Kavraki, and Thrun)Hutchinson, Kantor, Burgard, Kavraki, and Thrun)

GradingGrading

Participation: read readings, attend Participation: read readings, attend class prepared to discuss readingsclass prepared to discuss readings

Presentation(s): read and understand Presentation(s): read and understand research paper, and make 20 min research paper, and make 20 min PPT presentation to classPPT presentation to class

Semester-long projectSemester-long project

Semester ProjectSemester Project

Topic of your choosing, advised and Topic of your choosing, advised and approved by instructorapproved by instructor

Groups of 1-3 studentsGroups of 1-3 students ScheduleSchedule

• Proposal (Feb.)Proposal (Feb.)• Mid term report / discussion (March)Mid term report / discussion (March)• Final presentation (end of April)Final presentation (end of April)

Project IdeasProject Ideas

Robot chessRobot chess Finding and tracking people indoorsFinding and tracking people indoors UI for assistive robot armsUI for assistive robot arms Analysis of human observation data Analysis of human observation data (Prof. Yu)(Prof. Yu)

Outdoor vehicle navigation Outdoor vehicle navigation (Prof. Johnson)(Prof. Johnson)

Motion in social contexts Motion in social contexts (Profs. Scheutz and (Profs. Scheutz and Sabanovic)Sabanovic)

Class WebsiteClass Website

Please visit website for course policies, Please visit website for course policies, important announcements, and detailed important announcements, and detailed schedule:schedule:http://www.cs.indiana.edu/classes/b659-hauserk/http://www.cs.indiana.edu/classes/b659-hauserk/

Or from Or from http://www.cs.indiana.edu/~hauserk/http://www.cs.indiana.edu/~hauserk/, click , click ‘Teaching’‘Teaching’