1

Creating Robotic Creating Robotic BehavioursBehaviourswith ROBOLABwith ROBOLAB

Eric WangEric WangUniversity of Nevada, RenoUniversity of Nevada, Reno

Technology School of the FutureTechnology School of the FutureAdelaide, AustraliaAdelaide, Australia

March 25, 2004March 25, 2004

Tell me, and I’ll forget.Tell me, and I’ll forget.Show me, and I’ll remember.Show me, and I’ll remember.But involve me, and I’ll understand.But involve me, and I’ll understand.

--Chinese ProverbChinese Proverb

2

The Design ProcessThe Design Process

1.1. Identify the problem, including Identify the problem, including specifications and constraintsspecifications and constraints

2.2. Explore alternative solutionsExplore alternative solutions3.3. Design, build and test itDesign, build and test it4.4. Optimize and reOptimize and re--design the design the

solutionsolution5.5. Communicate and disseminate Communicate and disseminate

the solutionthe solution

Robotic BehaviorsRobotic Behaviors

•• BraitenbergBraitenberg Vehicles Vehicles •• Serial vs. behavioral programmingSerial vs. behavioral programming•• Schema theorySchema theory•• Knowing when you are stuckKnowing when you are stuck

3

BraitenbergBraitenberg VehiclesVehicles

•• Vehicles: Vehicles: Experiments into Experiments into Synthetic Synthetic PsychologyPsychology

•• Valentino Valentino BraitenbergBraitenberg

•• How hard is it to How hard is it to build (not build (not program) program) intelligence? intelligence?

Uphill Analysis, Downhill Uphill Analysis, Downhill InventionInvention

•• Behavioral observationsBehavioral observations•• The inferred properties are more The inferred properties are more

complicated than the known complicated than the known structurestructure

4

SimulationSimulation

QT version

SimulationSimulation

5

Vehicle 1: Getting AroundVehicle 1: Getting AroundExcitatory: motor

speed proportional to sensor output

Excitatory: motor Excitatory: motor speed proportional to speed proportional to sensor outputsensor output

Vehicle 1: Getting AroundVehicle 1: Getting Around

Is it ALIVE?Is it ALIVE?Is it ALIVE?

6

Vehicle 2: Fear & Vehicle 2: Fear & AggressionAggression

• Excitatory• 2 sensors• 2 motors• Crossed

•• ExcitatoryExcitatory•• 2 sensors2 sensors•• 2 motors2 motors•• CrossedCrossed

Vehicle 2: Fear & Vehicle 2: Fear & AggressionAggression

• “Aggression”•• ““Aggression”Aggression”

7

LEGO Vehicle 2: LEGO Vehicle 2: AggressionAggression

Vehicle 2: Fear & Vehicle 2: Fear & AggressionAggression

• Uncrossed • “Fear”•• Uncrossed Uncrossed •• “Fear”“Fear”

8

LEGO Vehicle 2: FearLEGO Vehicle 2: Fear

Vehicle 3: LoveVehicle 3: Love

• Inhibitory• Additional sensors•• InhibitoryInhibitory•• Additional sensorsAdditional sensors

9

Vehicle 3 combinationsVehicle 3 combinations

•• 2 sensor types (IR & visible)2 sensor types (IR & visible)•• 2 factors:2 factors:

–– Crossed or uncrossedCrossed or uncrossed–– Excitatory or inhibitoryExcitatory or inhibitory

•• 16 possible vehicle combinations16 possible vehicle combinations•• Still no programming!Still no programming!

LEGO sensorsLEGO sensors

CdS sensor with IR filter

LEGO light sensor with visible light filter

10

LEGO Vehicle 3: LoveLEGO Vehicle 3: Love

Advanced VehiclesAdvanced Vehicles•• 4: Value and Special Tastes4: Value and Special Tastes

–– Introduce nonlinear & discontinuous motorIntroduce nonlinear & discontinuous motor--sensor sensor relationshipsrelationships

–– e.g. saturation, e.g. saturation, hysteresishysteresis, etc., etc.–– Vehicles can make “decisions”Vehicles can make “decisions”

•• 5: Logic5: Logic–– Introduce logic gates & time delayIntroduce logic gates & time delay–– Crude memory?Crude memory?

•• 6: Selection, the Impersonal Engineer6: Selection, the Impersonal Engineer–– Introduce random perturbations in designIntroduce random perturbations in design–– Genetic algorithms?Genetic algorithms?

•• 7: Concepts7: Concepts–– Introduce “Introduce “MnemotrixMnemotrix” wire” wire–– Sensor response increases with each activationSensor response increases with each activation

11

ROBOLAB ModesROBOLAB Modes

Pilot ModePilot Mode

Inventor ModeInventor Mode

Investigator ModeInvestigator Mode

Float vs. StopFloat vs. Stop

Float StopFloat Stop(electric brake) (electric brake)

12

StructuresStructures

••JumpsJumps••LoopsLoops••ForksForks••TasksTasks

Simple loopsSimple loops

Loop 10 times (the default is to loop only twice!)Loop 10 times (the default is to loop only twice!)

13

Simple ControlSimple Control

Nested loopsNested loops

5, 10, 9, 14, 19, 18, 23, 28, 27, 32, 5, 10, 9, 14, 19, 18, 23, 28, 27, 32, 37, 36, 41, 46, 4537, 36, 41, 46, 45

14

LEGO cockroachLEGO cockroach

Basic forkBasic fork

Forks are equivalent to “If, then” Forks are equivalent to “If, then” statements (conditionals)statements (conditionals)

All forks must have a All forks must have a “merge”“merge”

15

Nested forksNested forks

Jumps & forksJumps & forks

Okay to jump between paths in a Okay to jump between paths in a forkfork

16

Containers & forksContainers & forks

Picking 2 different random Picking 2 different random numbersnumbers

17

Random ForkRandom Fork

ChallengeChallenge

Line followerLine follower1.1. Loops & wait Loops & wait forsfors2.2. Forks & loopsForks & loops3.3. Two light sensors, forks, & loopsTwo light sensors, forks, & loops•• Wall follower?Wall follower?

18

Line Follower #1Line Follower #1(start on right side of black line)(start on right side of black line)

Loops & Wait Loops & Wait ForsFors1.1. “Pivot” left until dark“Pivot” left until dark2.2. Pivot right until lightPivot right until light3.3. RepeatRepeat

Line Follower #2 Line Follower #2 (start on right side of black line)(start on right side of black line)

Forks & LoopsForks & Loops1.1. If light, pivot leftIf light, pivot left2.2. If dark, pivot rightIf dark, pivot right3.3. RepeatRepeat

19

Line Follower #3Line Follower #3(start straddling black line)(start straddling black line)

•• If left sensor dark, pivot rightIf left sensor dark, pivot right•• If right sensor dark, pivot leftIf right sensor dark, pivot left•• If neither dark, drive straight If neither dark, drive straight

Feeding FrenzyFeeding Frenzy

•• Serial vs. Behavioral ProgrammingSerial vs. Behavioral Programming•• Look for “food” with light sensorLook for “food” with light sensor

1.1. Search for food systematicallySearch for food systematically2.2. Search for food randomlySearch for food randomly

•• Beep & pause when eatingBeep & pause when eating•• Observe feeding pattern & amount Observe feeding pattern & amount

of food foundof food found

20

TasksTasks

••Task splitTask split••Stop tasksStop tasks••Start tasksStart tasks

Tasks are like Tasks are like programs.programs.

MutliMutli--tasking means tasking means you can run more you can run more than one program than one program at a time.at a time.

Task SplitTask Split

All tasks must end with their own stoplightAll tasks must end with their own stoplight

21

Resource conflictsResource conflicts

Both tasks Both tasks can access can access the sensor the sensor valuesvalues

Tasks will Tasks will conflict conflict over motor over motor controlcontrol

Sharing Control of MotorsSharing Control of Motors

Motor A goes forward for 2 sec, backwards Motor A goes forward for 2 sec, backwards for 2 sec and forward for 10 secfor 2 sec and forward for 10 sec

22

Tasks & ContainersTasks & Containers

User containers 32User containers 32--47 are “local” and are not shared between tasks47 are “local” and are not shared between tasks

Stopping tasksStopping tasks

Stopping all tasks is equivalent to ending Stopping all tasks is equivalent to ending the entire program. The sound is never the entire program. The sound is never played!played!

23

Stopping One TaskStopping One Task

Start TasksStart Tasks

The upper task can be restarted by The upper task can be restarted by the lower taskthe lower task

24

Task PriorityTask Priority

Putting it all togetherPutting it all together

25

Getting StuckGetting Stuck

Realizing You Are StuckRealizing You Are Stuck

26

Random Motion OccasionallyRandom Motion Occasionally

EventsEvents

27

What is an Event?What is an Event?

•• An An EventEvent is like a combination of a is like a combination of a Wait ForWait For and a and a JumpJump..

Types of EventsTypes of Events

•• 16 Event conditions and 3 Event 16 Event conditions and 3 Event ColorsColors

28

Setting up an EventSetting up an Event

Stopping & ReStopping & Re--starting starting EventsEvents

•• You can stop monitoring an event and You can stop monitoring an event and then restart it laterthen restart it later

29

Restarting EventsRestarting Events

•• Restarting can be after the Event Restarting can be after the Event LandingLanding

•• Program will jump backwards on the Program will jump backwards on the EventEvent

Multiple EventsMultiple Events

•• 3 standard color Events3 standard color Events•• Up to 16 EventsUp to 16 Events•• All Events share same Event LandingAll Events share same Event Landing

30

Multiple Events & TasksMultiple Events & Tasks

•• One Event Landing per TaskOne Event Landing per Task•• 16 total Events for all Tasks16 total Events for all Tasks

Obstacle avoidance using Obstacle avoidance using only the light sensoronly the light sensor

31

Communication: Communication: handshakinghandshaking

•• Cannot send & receive Mail at same Cannot send & receive Mail at same timetime

•• Need to work out handshaking Need to work out handshaking protocolprotocol

ChallengeChallenge

•• Build a robot that wanders the Build a robot that wanders the room (randomly or systematically) room (randomly or systematically) using touch and/or light sensorsusing touch and/or light sensors

•• Use Events to detect if you might Use Events to detect if you might be stuck (too much time has be stuck (too much time has elapsed since anything has elapsed since anything has happened)happened)

•• Call for help if you are stuckCall for help if you are stuck

32

CameraCamera

Vision Center

33

For More InformationFor More Information

ROBOLAB support: ROBOLAB support: •• www.ceeo.tufts.eduwww.ceeo.tufts.edu

Email: Email: •• eric.wang@unr.edueric.wang@unr.edu

THE ENDTHE END

34