Adaptive Robotics

Adaptive Robotics

COM2110Autumn Semester 2008Lecturer: Amanda Sharkey

“Robot”

the word “robot” comes from the play `Rossum`s Universal Robots`, by Czech writer Karel Capek (1921)

Robot, from robota, “servitude, forced labour, drudgery”

Robots rebel, and kill all humans

What is a robot?

Joseph Engelberger, a pioneer in industrial robotics: "I can't define a robot, but I know one when I see one."

Brady (1985)

“the intelligent connection of perception to action”

Arkin (1998)“An intelligent robot is a machine able

to extract information from its environment and use knowledge about its world to move safely in a meaningful and purposive manner”

Robotics Industry Association:“a robot is a re-programmable, multi-functional, manipulator designed to move material, parts, tools or specialised devices through variable programmed motions for the performance of a variety of tasks”

(excludes mobile robots!)

Changing definitions “Stop fearing the robot – stop making a man of him! Just

remember that the sewing machine is a robot, the automobile is a robot, the electric car and the phonograph and the telephone are all robots. Each one men have developed in order to unburden themselves of some onerous task and on to better things. Each one does a specific job, and no more. Why begin now to worry about robots when we have been enjoying their

services for centuries?” Woodbury, D. (1927) Dramatising the

“robot”, New York Times, Nov 1st

Like Wittgenstein and “games” No single feature shared by the many

examples, but rather “a complicated network of similarities, overlapping and criss-crossing” [Wittgenstein, 1953].

The same is also true of ‘robot’ – the various examples bear family resemblances rather than a single meaning.

Different groups of robots Autonomous robots Industrial robots Human-like robots Self-configurable robots Biological models Toys and companions

Course Aims To present the key concepts of a recent

approach to AI And contrast to earlier approaches

To consider the underlying mechanisms for robot control To inform about research in robotics

What are the motivations? Applications Biological inspiration Biorobotic modelling Understanding intelligence

Teaching Method

Lectures, and assignment.See website for course (Lecturer’s

module pages)

Assessment: Exam and assignment

Background ReadingClark, A. (1997) Being There: Putting Brain, Body

and World Together Again. A Bradford Book, MIT Press

Franklin, S. (1995) Artificial Minds: A Bradford Book, MIT Press

Nolfi, S. and Floreano, D. (2000) Evolutionary Robotics: The biology, intelligence and technology of self-organising machines. A Bradford Book, MIT Press

Pfeifer, R., and Scheier, C. (2001) Understanding Intelligence, MIT Press

Why robotics?

Can we create artificial beings? Are we machines? How do we work? Understanding by building

Making robots to perform useful tasks Robots as companions?

What is Adaptive Robotics?Recent approach to AI

Reflected in • Behaviour based robotics• Reactive robotics• Evolutionary robotics• Artificial Life• Swarm Intelligence and swarm robotics• Embodied cognition

Different views of mind and cognition

Emphasis on Mind and Reasoning independent of world (computationalism)

How can mind emerge from the workings of a physical machine? (brain) (connectionism)

Relationship between brain, body, mind and world…. (embodied cognition)

Three Stage Progression to current emphasis on Embodied Cognition

1. Classical Cognitivism or computationalism (late 1950’s to 1980’s)

2. Connectionism (main period– 1980’s)3. Embodied Cognition and Adaptive

Intelligence (1990’s to present)

N.B. dates only a rough guide

1. Computationalism

Mental states = computational statesGood Old Fashioned Artificial Intelligence GOFAI

Physical Symbol System Hypothesis (Newell and Simon, 1976)

A physical symbol system is a necessary and sufficient condition for general intelligent action.

intelligence is symbol manipulationcomputers manipulate symbols

computers can be intelligent

1. Computationalism cont. Memory as retrieval from stored

symbolic database Problem solving as logical inference Cognition as centralised Environment just a problem domain Body as an input device

Shakey

Functionalism“The mind is to the brain as the

program is to the hardware” (Johnson-Laird, 1988)

- hardware/software distinction- we are interested in the software –

could run on any hardware (Swiss cheese?)

2. Connectionism Neural nets An account of mental states in terms

of neurons – related to brain Memory as pattern recreation Problem solving as pattern

completion and transformation Cognition – decentralised

3. Embodied Cognition

As connectionism PLUS

Environment as active resource Body as part of computational loop

Brain, body, world intricately interconnected

3. Embodied cognition cont. Gradual move away from

anthropocentric view Greater awareness of abilities of

non-human organisms, and their abilities to interact with and survive in the world.

Early mobile robots: Shakey

Shakey the Robot Developed by SRI (Stanford Research

Institute) from 1966-1972 First mobile robot to visually interpret,

and reason about its surroundings TV camera, range finder, bump sensors Programs for sensing, modelling and

planning Example task: “push the block off the

platform”

Stanford Cart TV cameras: took pictures of scenes,

and planned path between obstacles

Sense Model Plan Action

Brooks:1991“Intelligence without representation”

Realisation that mobility, vision and ability to survive are important aspects of intelligence

Brooks and idea of Creatures Able to cope with changing and

uncertain world Should have goals, and purpose in

being

“An ant, viewed as a behaving system, is quite simple. The apparent complexity of its behavior over time is largely a reflection of the complexity of the environment in which it finds itself”

Herbert A. Simon, 1969

Idea of reactive responses to the world, instead of modelling and planning.

Intelligence is determined by the dynamics of interaction with the world.

Key concepts in new approach to AI a) Reactive behaviour b) Adaptivity c) Situatedness d) Embodiment e) Emergence and Self-organisation Changing view of intelligence

a. Reactive Intelligence

Arkin (1995): hallmark characteristics- emphasis on behaviours and simple

sensorimotor pairings- Avoidance of abstract representational

knowledge (time consuming)- Animal models of behaviour- Demonstrable results: walking robots,

pipe-crawling robots, military robots etc.

Reactivity Biological inspiration: e.g. birds flocking, ants foraging. Sufficiency

Grey Walter (1953) electronic tortoise. Braitenberg (1984) synthetic psychologyBrooks (1986) behaviour-based robotics

and subsumption architecture.

b. Adaptivity Adaptivity: ability to adjust oneself to the

environment Physiological adaptation – e.g. sweating to

adjust to heat Evolutionary adaptation – e.g. peppered

moth. Light in colour, in industrial area became dark in colour

Sensory adaptation – e.g. our pupils adjusting to poor light

Adaptation by learning – e.g. where food is found

c. Situated

An emphasis on robot’s interaction with its environment (related to embodiment)

Brooks (1991) “the world is its own best model”

A situated agent must respond in a timely fashion to its inputs.

d. Embodiment Physical grounding of robot in the worldBrooks (1991): embodiment of intelligent

systems critical because Only an embodied intelligent agent is fully

validated as one that can deal with the real world

Only through physical grounding can any meaning be given to the processing occurring within the agent

“Intelligence is determined by the dynamics of interaction with the world” (Brooks 1991)

- embodied cognition- A solution to the symbol grounding

problem?- (remember Searle’s Chinese Room!)

e. Emergence Adaptive success that emerges from

complex interactions between body, world and brain

A non-centrally controlled (or designed) behaviour that results from the interactions of multiple simple components

Meanings of the term ‘emergence’ Surprising situations or behaviours Property of system not contained in

any of its parts Behaviour resulting from agent-

environment interaction that is not explicitly programmed.

Ant colony Individual ants are simple and

reactive (?) Emergent behaviour of colony is

sophisticated

Self-organisationAn ant colony is self-organised – simple

individuals, local interactions, emergent behaviour .. No global control

“self-organisation is a set of dynamical mechanisms whereby structures appear at the global level of a system from interactions among its lower-level components. The rules specifying the interactions among the system’s constituent units are executed on the basis of purely local information, without reference to the global pattern, which is an emergent property of the system rather than a property imposed upon the system by an external ordering influence”

(Bonabeau, Dorigo and Theraulaz, 1999)

Frisbee collecting robots Robots in an arena + frisbees Simple rules Emergent result – clustering and

sorting of frisbees.

Changing view of intelligence

GOFAI – emphasis on reasoning, planning, and representation. Human-centred (anthropocentric)

Behaviour-based robotics and beyond: emphasis on simpler organisms and their ability to survive in the world.

Reading: For this week and next.

Brooks, R.A. (1991) Intelligence without Reason. Proceedings of 1991 International Joint Conference on Artificial Intelligence, 569-595.

Robots in the news

Murata Manufacturing: Murata boy – controlled by blue tooth, and can ride a bike forwards and backwards.

9/25/2008 12:23 PM ET iRobot Corp. (IRBT: News ), on Thursday, said that it has received an additional $13.3 million order from the US Army for PackBot 510 with FasTac Kit robots for carrying bomb identification and other life-threatening missions.