+ All Categories
Home > Documents > Robot Intelligence

Robot Intelligence

Date post: 02-Jan-2016
Category:
Upload: david-espinoza
View: 28 times
Download: 1 times
Share this document with a friend
Description:
Robot Intelligence. Kevin Warwick. Task Example. We will look at a relatively simple example and see how a robot might solve a specific problem Try to spot the assumptions we/humans make on the robot’s capabilities - PowerPoint PPT Presentation
24
Robot Intelligence Robot Intelligence Kevin Warwick Kevin Warwick
Transcript
Page 1: Robot Intelligence

Robot IntelligenceRobot Intelligence

Kevin Warwick Kevin Warwick

Page 2: Robot Intelligence

Task ExampleTask Example

• We will look at a relatively simple We will look at a relatively simple example and see how a robot might example and see how a robot might solve a specific problemsolve a specific problem

• Try to spot the assumptions we/humans Try to spot the assumptions we/humans make on the robot’s capabilitiesmake on the robot’s capabilities

• In the problem the robot will start at In the problem the robot will start at one point and must find it’s way to the one point and must find it’s way to the goalgoal

Page 3: Robot Intelligence

Topological Representation Topological Representation

Page 4: Robot Intelligence

Topological RepresentationTopological Representation

• The space can The space can be partitioned be partitioned using for using for example:example:

• Binary State Binary State Partitioning.Partitioning.

• But does the But does the robot have or robot have or can it find out can it find out such a map?such a map?

Page 5: Robot Intelligence

A connectivity graph of this A connectivity graph of this BSP is:BSP is:

Page 6: Robot Intelligence

Both together…Both together…

Page 7: Robot Intelligence

Valid RoutesValid Routes

• In order to get from the start position (region 2) to the goal (region In order to get from the start position (region 2) to the goal (region 14) the robot must follow one of the routes from 14) the robot must follow one of the routes from 22 to to 1414 in the in the graph graph

• 2 2 3 3 4 4 5 5 6 6 10 10 13 13 12 12 14 14• 2 2 3 3 4 4 9 9 10 10 13 13 12 12 14 14• 2 2 3 3 4 4 8 8 7 7 11 11 12 12 14 14• 2 2 1 1 7 7 11 11 12 12 14 14 • 2 2 1 1 7 7 8 8 4 4 5 5 6 6 10 10 13 13 12 12 14 14• 2 2 1 1 7 7 8 8 4 4 9 9 10 10 13 13 12 12 14 14

Page 8: Robot Intelligence

Real-time Solutions in Real-time Solutions in Unmodelled EnvironmentsUnmodelled Environments

• If the robot did not know the If the robot did not know the structure of the environment and structure of the environment and could not build a planned trajectory, could not build a planned trajectory, the goal may still be reachedthe goal may still be reached

• Consider the goal having a beacon to Consider the goal having a beacon to indicate to the robot where the goal indicate to the robot where the goal isis

Page 9: Robot Intelligence

Real-time Solutions in Real-time Solutions in Unmodelled Environments Unmodelled Environments

• Now consider a wall following routineNow consider a wall following routine– The robot moves forward following the The robot moves forward following the

left hand wall left hand wall until until • it reaches an obstacle (a wall in front)it reaches an obstacle (a wall in front)

– It turns right – possibly repeatedlyIt turns right – possibly repeatedly

• it no longer detects the presence of a wall to it no longer detects the presence of a wall to the leftthe left

– It turns leftIt turns left

– Can the robot reach the goal?Can the robot reach the goal?

Page 10: Robot Intelligence

Problem RevisitedProblem Revisited

Page 11: Robot Intelligence

Real-time Solutions in Real-time Solutions in Unmodelled EnvironmentsUnmodelled Environments

• In this case the robot will be able to In this case the robot will be able to reach the goal if the robot initially reach the goal if the robot initially moves to the rightmoves to the right– If the robot moves to the left then it will If the robot moves to the left then it will

circle the top-left island and would require circle the top-left island and would require some additional reasoning to break out of some additional reasoning to break out of that cyclethat cycle•e.g. Odometry would indicate that the pose of e.g. Odometry would indicate that the pose of

the robot has repeated a number of timesthe robot has repeated a number of times

Page 12: Robot Intelligence

Robot ArchitecturesRobot Architectures

• Example: seven dwarfExample: seven dwarf– Pseudocode …Pseudocode …

• if if left_sensor_reading left_sensor_reading – turn turn rightright

•elseif elseif right_sensor_readingright_sensor_reading – turn turn leftleft

•elseif elseif ((right_sensor_readingright_sensor_reading) ) and and ((left_sensor_readingleft_sensor_reading))

– reversereverse

•elseelse– move forwardmove forward

Page 13: Robot Intelligence
Page 14: Robot Intelligence

Robot ArchitecturesRobot Architectures

– Add in wall following …Add in wall following …• if if left_sensor_reading left_sensor_reading

– if if left_sensor_readingleft_sensor_reading < min < min turn turn right_slightlyright_slightly

– if if left_sensor_readingleft_sensor_reading > max > max turn turn left_slightlyleft_slightly

– elseelse move forwardmove forward

• elseif elseif right_sensor_readingright_sensor_reading – if if right_sensor_readingright_sensor_reading < min < min

turn turn left_slightlyleft_slightly– if if right_sensor_readingright_sensor_reading > max > max

turn turn right_slightlyright_slightly– elseelse

move forwardmove forward• elseif elseif ((right_sensor_readingright_sensor_reading) ) and and ((left_sensor_readingleft_sensor_reading))

– reversereverse• elseelse

– move forwardmove forward

• Even introducing modest extensions can lead to increased complexity Even introducing modest extensions can lead to increased complexity in the algorithm if it is developed in an in the algorithm if it is developed in an ad hocad hoc way way

• Need to find some more formalised way of developing behavioursNeed to find some more formalised way of developing behaviours

Page 15: Robot Intelligence

Robot ArchitecturesRobot Architectures

• Two main classesTwo main classes– CentralisedCentralised– ReactiveReactive

• Three main types:Three main types:– High Level Control (centralised)High Level Control (centralised)

• Treat the robot as an abstract entityTreat the robot as an abstract entity• Apply classical AI techniques to define complex tasksApply classical AI techniques to define complex tasks• Top down (Computer Science) approachTop down (Computer Science) approach

– Functional (centralised)Functional (centralised)• Classical horizontal connection between perception and Classical horizontal connection between perception and

actionaction• sense – model – plan – act sense – model – plan – act paradigmparadigm

Page 16: Robot Intelligence

Robot ArchitecturesRobot Architectures

– Reactive (distributed)Reactive (distributed)• Bottom up (Cybernetics) approachBottom up (Cybernetics) approach• Subsumption Subsumption

– most widely knownmost widely known

• Motor Schema Motor Schema • Ego-behaviour Ego-behaviour

• Biological AnaloguesBiological Analogues– Artificial Neural NetworksArtificial Neural Networks– Genetic AlgorithmsGenetic Algorithms

• Hybrid systemsHybrid systems– Combinations of any/all of the aboveCombinations of any/all of the above

Page 17: Robot Intelligence

Centralised Controller Centralised Controller (functional)(functional)

Page 18: Robot Intelligence

Distributed Controller Distributed Controller (reactive)(reactive)

Page 19: Robot Intelligence

Functional Architectures I: Functional Architectures I: HierarchicalHierarchical• Decompose the control process by functionDecompose the control process by function• Low level processes provide simple functions that are grouped Low level processes provide simple functions that are grouped

together to provide higher level functionstogether to provide higher level functions• e.g.e.g.

– Lower level processesLower level processes• position sensingposition sensing• motor outputmotor output• forward kinematicsforward kinematics• inverse kinematicsinverse kinematics• dynamic model of the robotdynamic model of the robot• low level vision processing (e.g. edge detection)low level vision processing (e.g. edge detection)• ……• Note:Note: while these are “low level” some of them can be computationally while these are “low level” some of them can be computationally

challenging, especially kinematics, dynamics and visionchallenging, especially kinematics, dynamics and vision– Higher level processesHigher level processes

• Mission planningMission planning• Map buildingMap building• Reasoning about tasksReasoning about tasks• Sequencing tasksSequencing tasks

Page 20: Robot Intelligence

Functional Architectures II: Functional Architectures II: BlackboardBlackboard• Blackboard-based architectures rely on a common pool of information Blackboard-based architectures rely on a common pool of information

(the blackboard) that is shared by independent computational (the blackboard) that is shared by independent computational processesprocesses

• Example: Ground Surveillance Robot (GSR)Example: Ground Surveillance Robot (GSR)– designed to navigate from one designed to navigate from one knownknown location to another location to another knownknown location location

over over unknownunknown natural terrain natural terrain– to complete the task the GSR has to build a terrain mapto complete the task the GSR has to build a terrain map– GSR uses the blackboard to represent and pass information from one GSR uses the blackboard to represent and pass information from one

software module to anothersoftware module to another– there is a loose coupling between modulesthere is a loose coupling between modules

• DisadvantagesDisadvantages– requires the information to be consistent in its representation requires the information to be consistent in its representation

• cannot use relative positions as defined in one subsystem in anothercannot use relative positions as defined in one subsystem in another– can lead to bottlenecks in processingcan lead to bottlenecks in processing– asynchronous nature of blackboards can lead to problems due to timing asynchronous nature of blackboards can lead to problems due to timing

skewsskews– difficulties arise if more than one module can change data in the pool … difficulties arise if more than one module can change data in the pool …

which set of data is the right one?which set of data is the right one?• similar problem to file sharing violationssimilar problem to file sharing violations

Page 21: Robot Intelligence

Example Blackboard SystemExample Blackboard System

Page 22: Robot Intelligence

Basic ReactiveBasic Reactive

• Consider instead a very basic reactive Consider instead a very basic reactive robot:robot:

• Higher level goal – move towards a Higher level goal – move towards a desired end position (infra-red?)desired end position (infra-red?)

• Lower level behaviour – move Lower level behaviour – move forwardsforwards

• Lower level behaviour – avoid Lower level behaviour – avoid obstaclesobstacles

Page 23: Robot Intelligence

Basic ReactiveBasic Reactive• Lower level behaviours – exampleLower level behaviours – example• Kohonen network to map robot positional Kohonen network to map robot positional

state/situation (object to left/right/front – state/situation (object to left/right/front – distance)distance)

• Fuzzy Automata – Physical realisation in Fuzzy Automata – Physical realisation in each state – e.g. left wheel forward fast, each state – e.g. left wheel forward fast, right wheel reverse slowlyright wheel reverse slowly

• Needs reward/punishmentNeeds reward/punishment• Probabilities randomised then updated Probabilities randomised then updated

based on success/failure of moving based on success/failure of moving forward and avoiding obstacles forward and avoiding obstacles

Page 24: Robot Intelligence

Next PresentationNext Presentation

• ArchitecturesArchitectures


Recommended