15-491 – CMRoboBits: Creating Intelligent Robots

Introduction

Instructor: Manuela VelosoFall 2008

TAs: Stephanie Rosenthal and Richard Wanghttp://www.andrew.cmu.edu/course/15-491

Computer Science DepartmentCarnegie Mellon

15-491 CMRoboBits

What is a Robot?

Many robots…

15-491 CMRoboBits

15-491 CMRoboBits

15-491 CMRoboBits

15-491 CMRoboBits

AIBO ERS-7

CCD Camera208x160 pixels

Microphones

LEDs

IR Proximity Sensor

Foot Pads

3-Axis Accelerometers

Wifi Card

576MHz MIPS CPU

15-491 CMRoboBits

What is a Robot?

15-491 CMRoboBits

15-491 CMRoboBits

Intelligent Complete Robot

Action

Actuators

Perception

External WorldExternal World

Sensors

Cognition

Sensing, modeling the world

Behaviors, action selection, planning, learningMulti-robot coordination, teamwork

Response to opponent, multi-agent learning

Motion, navigation, obstacle avoidance

15-491 CMRoboBits

Course Motivations

Robots are integrated intelligence:Sensing & perceptionBehavior & cognitionLearning from the real worldMotion & kinematicsMulti-robot cooperation & coordination

15-491 CMRoboBits

15-491 CMRoboBits

The CMRoboBits Course

Making robots accessible to allHow do you solve problems in “real-time”?Sensors are inherently noisy. How do you develop behaviors robust to errors?How do you coordinate the actions of multiple robots?

15-491 CMRoboBits

CMRoboBits Fall 2008

Multiple robotsiRobot Create, www.irobot.comScribbler, www.scribblerrobot.com

15-491 CMRoboBits

Centralized Perception, Centralized Control

Robot team is autonomous as a whole

Wireless Link

Global vision

Offboardcomputation

10 robots + golf ball 5 robots per team

Color markers for ID

15-491 CMRoboBits

Scribbler, Simplest Obstacle Avoidancedef main():

while True:L,R = getIR()

if L and R:move(0,0)

elif L:move(0.6,-0.5)

elif R:move(0.6,0.5)

else:move(0.6,0)

15-491 CMRoboBits

CMRoboBits Main Topics

MotionSensing BehaviorsMulti-robot coordination

15-491 CMRoboBits

Actions

Effector – any device to effect on the environment

Legs, wheels, arms, fingers, speakers, etcPhysical work of some kind – walking, talking

Actuator – a mechanism that enables the effector

15-491 CMRoboBits

Passive Actuation (e.g. Delft, MIT, Cornell)

15-491 CMRoboBits

Active ActuationElectric motors

Most commonHydraulics

Fluid pressure changes, the actuator movesPowerful, precise, large, lots of care

PneumaticsAir pressure

Photo-reactive materialsReact to light, small work, micro-robotics

Chemically, thermally, … reactive materials

15-491 CMRoboBits

ControlHolonomic

Controllable DOF = TDOFHelicopter (6 DOFs)

NonholonomicCDOF < TDOFCar 2/3, parallel parking

RedundantCDOF > TDOFHuman arm Many solutions to moving from one place to other

15-491 CMRoboBits

Uncertainty of Robot ActionsExample: AIBO kicks as planned motions

frame-based with kinematic interpolation

15-491 CMRoboBits

Handling Uncertainty: Model-Based Kick Selection

Modeling effects of different kicksAngle analysis

Strength analysis

-100 -80 -60 -40 -20 0 20 40 60 80 1000

5

10

15

20

25

Left Head Forward Right Head

-4000 -3500 -3000 -2500 -2000 -1500 -1000 -500 0 500 1000-3000

-2000

-1000

0

1000

2000

3000Normal Head Kick

-4000 -3500 -3000 -2500 -2000 -1500 -1000 -500 0 500 1000-3000

-2000

-1000

0

1000

2000

3000Hard Head Kick

15-491 CMRoboBits

Sensors

Bump sensorsAccelerometerVision

15-491 CMRoboBits

Cognition: Behaviors

Score Search

Approach

Recovernot see ball

next to goaland ball

see ball not see ball

timeout

see ballnot next to ball

15-491 CMRoboBits

Conclusion

Working within the perception, cognition, action loop Working with multiple robotsMicrosoft Robotics Studio – MSRS

An exciting course.

www.andrew.cmu.edu/course/15-491