PCT
`
`WORLD INTELLECTUAL PROPERTY ORGANIZATION
`International Bureau
`
`
`
`INTERNATIONAL APPLICATION PUBLISHED UNDER THE PATENT COOPERATION TREATY (PCT)
`
`(51) International Patent Classification 6;
`
`(11) International Publication Number:
`
`WO 00/04430
`
`27 January 2000 (27.01.00)
`
`
`
`
`
`GO0SD 1/02
`(43) International Publication Date:
`
`
`
`PCT/US99/16078|(81) Designated States: AE, AL, AM, AT, AT (Utility model), AU,
`(21) International Application Number:
`AZ, BA, BB, BG, BR, BY, CA, CH, CN, CU, CZ, CZ
`
`
`
`(Utility model), DE, DE (Utility model), DK, DK (Utility
`(22) International Filing Date:
`16 July 1999 (16.07.99)
`model), EE, EE (Utility model), ES, FI, FI (Utility model),
`
`GB, GD, GE, GH, GM, HR, HU, ID,IL,IN, IS, JP, KE,
`
`
`KG, KP, KR, KZ, LC, LK, LR, LS, LT, LU, LV, MD, MG,
`(30) Priority Data:
`
`MK, MN, MW, MX, NO, NZ, PL, PT, RO, RU, SD, SE,
`98305761.3
`20 July 1998 (20.07.98)
`EP
`SG, SI, SK, SK (Utility model), SL, TJ, TM, TR, TT, UA,
`
`UG, US, UZ, VN, YU, ZA, ZW, ARIPO patent (GH, GM,
`
`
`KE, LS, MW,SD,SL, SZ, UG, ZW), Eurasian patent (AM,
`(71) Applicant (for all designated States except US): THE PROC-
`
`AZ, BY, KG, KZ, MD, RU, TJ, TM), European patent (AT,
`TER & GAMBLE COMPANY [US/US]; One Procter &
`
`BE, CH, CY, DE, DK, ES, FI, FR, GB, GR, IE, IT, LU,
`Gamble Plaza, Cincinnati, OH 45202 (US).
`MC,NL, PT, SE), OAPI patent (BF, BJ, CF, CG, CI, CM,
`
`GA, GN, GW, ML, MR, NE, SN, TD, TG).
`(72) Inventors; and
`
`(75) Inventors/Applicants (for US only):
`BOTTOMLEY,
`Ian
`
`
`[GB/GB];
`17 Howick Street, Alnwick, Northumberland
`NE66 1UZ (GB). COATES, David [GB/GB]; 14 Percy
`
`
`Road, Shilbottle, Northumberland NE66 2HF (GB). GRAY-
`DON, Andrew, Russell [GB/GB]; 42 The Wills Building,
`Wills Oval, Newcastle-upon—-Tyne NE7 7RW (GB).
`
`
`
`(74) Agents: REED, T., David et al.; The Procter & Gamble
`Company, 5299 Spring Grove Avenue, Cincinnati, OH
`45217-1087 (US).
`
`Published
`With international search report.
`
`
`
`(54) Title: ROBOTIC SYSTEM
`
`(57) Abstract
`
`A self-propelled robot is disclosed for movement over a surface to be treated. The robot has a power supply (11) and a pair of
`wheels (8,9) driven by motors (6, 7) for moving the robot over the ssurface. A mechanism (113, 115, 16) is provided for controllably
`depositing a fluent material onto the surface. Navigation sensors (4, 13, 18, 21) provide signals for enabling the robot to navigate over
`the surface and one or more detectors (14, 15, 17) detect the presence of the material on the surface and provide signals indicative ofits
`presence. A control system (100) receives the signals from the sensors and detectors and controls the motors and the depositing mechanism
`in dependence upon the signals received from the sensors and detectors.
`
`Silver Star Exhibit 1006
`
`Silver Star Exhibit 1006
`
`

`

` AL
`
`AM
`AT
`AU
`AZ
`BA
`BB
`BE
`BF
`BG
`BJ
`BR
`BY
`CA
`CF
`CG
`CH
`cI
`«M
`CN
`cu
`CZ
`DE
`DK
`EE
`
`Codes used to identify States party to the PCT on the front pages of pamphlets publishing international applications under the PCT.
`Slovenia
`SI
`Lesotho
`LS
`Slovakia
`LT
`SK
`Lithuania
`SN
`LU
`Senegal
`Luxembourg
`Swaziland
`SZ
`Latvia
`LV
`TD
`Chad
`MC
`Monaco
`TG
`MD
`Togo
`Republic of Moldova
`MG
`TJ
`Tajikistan
`Madagascar
`TM
`Turkmenistan
`MK
`The former Yugoslav
`TR
`Turkey
`Republic of Macedonia
`TT
`Mali
`Trinidad and Tobago
`UA
`Ukraine
`Mongolia
`UG
`Mauritania
`Uganda
`US
`United States of America
`Malawi
`UZ
`Uzbekistan
`Mexico
`VN
`Viet Nam
`Niger
`YU
`Netherlands
`Yugoslavia
`ZW
`Zimbabwe
`Norway
`New Zealand
`Poland
`Portugal
`Romania
`Russian Federation
`Sudan
`Sweden
`Singapore
`
`Albania
`Armenia
`Austria
`Australia
`Azerbaijan
`Bosnia and Herzegovina
`Barbados
`Belgium
`Burkina Faso
`Bulgaria
`Benin
`Brazil
`Belarus
`Canada
`Central African Republic
`Congo
`Switzerland
`Céte d’Ivoire
`Cameroon
`China
`Cuba
`Czech Republic
`Germany
`Denmark
`Estonia
`
`ES
`FI
`FR
`GA
`GB
`GE
`GH
`GN
`GR
`HU
`TE
`IL
`IS
`IT
`JP
`KE
`KG
`KP
`
`KR
`KZ
`LC
`LI
`LK
`LR
`
`FOR THE PURPOSES OF INFORMATION ONLY
`
`Spain
`Finland
`France
`Gabon
`United Kingdom
`Georgia
`Ghana
`Guinea
`Greece
`Hungary
`Treland
`Israel
`Iceland
`Italy
`Japan
`Kenya
`Kyrgyzstan
`Democratic People’s
`Republic of Korea
`Republic of Korea
`Kazakstan
`Saint Lucia
`Liechtenstein
`Sri Lanka
`Liberia
`
`
`
`Silver Star Exhibit 1006 - 2
`
`Silver Star Exhibit 1006 - 2
`
`

`

`WO 00/04430
`
`PCT/US99/16078
`
`ROBOTIC SYSTEM
`
`The present
`invention relates to robotic systems and,
`more particularly to a mobile robotic system capable of
`movement
`over
`a
`surface
`and
`capable
`of
`treating the
`
`surface.
`
`this
`robots, of
`or
`Conventionally robotic systems,
`self-
`type may
`be described as
`semi-autonomous,
`i.e.
`propelling
`but
`relying
`for
`navigational
`guidance
`on
`transmitters,
`receivers
`and
`sensors
`to
`establish
`a
`
`coordinate system by which the robot navigates,
`
`in effect
`
`learning the location of obstacles within its field of
`
`movement. More recently it has been proposed to allow a
`robot
`to move without establishing a coordinate system,
`instead relying on the sensing of ad hoc stimuli to enable
`the robot
`to navigate around obstacles.
`For example,
`it
`has been proposed to provide a
`robotic vacuum cleaner
`operating along these lines.
`Self-navigational
`robotic
`systems of this type are referred to as autonomous robots.
`
`However,
`
`robots of
`
`these types, often intended for
`
`operation in a domestic environment, need a control system
`which is capable of allowing the robot
`to move around its
`environment
`in safety and therefore additionally need some
`sort of collision detection system which is capable of
`providing information on collisions or impending collisions
`to a control
`system capable of acting very quickly to
`
`and
`prevent the collision or else to minimise the impact,
`to perform collision avoidance by re-orienting the robot
`
`before
`
`further
`
`movement.
`
`Unfortunately,
`
`on-board
`
`is inevitably limited by cost constraints
`processing power
`in particular and therefore present systems,
`to avoid be
`prohibitively
`expensive,
`have
`relatively
`limiting
`navigational abilities which result,
`in use,
`in the robot
`
`the same areas
`tracing a path which involves passing over
`of the surface on plural occasions. Whilst this may not be
`problematic in say a vacuum cleaner,
`if the robot has the
`
`Silver Star Exhibit 1006 - 3
`
`Silver Star Exhibit 1006 - 3
`
`

`

`WO 00/04430
`
`PCT/US99/16078
`
`then such
`function of treating the surface in other ways,
`redundant movement may
`result
`in over-treatment of
`the
`surface which is not only wasteful of the product used for
`the treatment
`(a
`serious problem where
`the payload is
`restricted), but may also damage the surface or otherwise
`
`actually be harmful.
`The present
`invention is aimed at providing a self-
`
`propelled robot which can overcome such problems.
`According to the present
`invention,
`there is provided
`
`a self-propelled robot for movement over a surface to be
`
`treated,
`
`the robot comprising
`
`a power supply;
`
`from the power
`a traction mechanism receiving power
`supply, for moving the robot over the surface;
`a mechanism for
`controllably depositing a
`
`fluent
`
`material on to the surface;
`
`a plurality of navigation sensors providing signals
`for enabling the robot to navigate over the surface;
`
`one or more detectors adapted to detect
`
`the presence
`
`of
`
`the material
`
`on
`
`the
`
`surface
`
`and
`
`provide’
`
`signals
`
`indicative thereof; and
`signals
`a
`control
`system receiving the
`sensors
`and
`detectors,
`for
`controlling
`the
`
`from the
`traction
`
`mechanism and the depositing mechanism in dependence upon
`
`the signals received from the sensors and detectors.
`By detecting the application of
`the fluent material,
`which may be a liquid or gaseous fluid or else a flowable
`powder,
`the over-application of material can be avoided or
`
`areas
`around
`minimised by either navigating the robot
`already treated and/or
`by
`controlling the
`depositing
`mechanism to stop the deposit
`of material
`over
`such
`previously treated areas.
`is preferably contained within
`Material for treatment
`and may
`comprise
`suitable
`a
`reservoir
`on
`the
`robot
`floors,
`compositions
`for
`treatment of
`carpets
`and other
`floor coverings.
`The robot may,
`if desired, also include
`
`Silver Star Exhibit 1006 - 4
`
`Silver Star Exhibit 1006 - 4
`
`

`

`WO 00/04430
`
`PCT/US99/16078
`
`floor covering prior to
`for cleaning the floor or
`means
`treatment,
`for example in the form of
`a vacuum cleaning
`
`device.
`
`The
`
`invention also includes a method of
`
`treating a
`
`surface using a
`
`robot as defined above.
`
`The
`
`treatment
`
`method may be used for various applications on carpets, and
`
`
`
`
`
`other as_cleaning,floor coverings, such protective
`treatment,
`for example for stain and soil protection, fire
`protection,
`UV protection, wear
`resistance,
`dust mite
`control, anti microbial
`treatment and the like, as well as
`treatment
`to
`provide
`an
`aesthetic
`benefit
`such
`as
`
`The treatment method may also
`odorization/deodorization.
`find application on other surfaces such as synthetic floor
`coverings,
`ceramics or wood.
`As well as polishing hard
`surfaces,
`the robot may also be used to apply coatings to
`either enhance aesthetics or to act as a protective layer.
`Thus, according to a further aspect of the invention,
`there is provided a method for controllably depositing a
`fluent material
`on
`to floors,
`carpets
`and other
`floor
`coverings using an autonomous, self propelled, deposition-
`sensing robot.
`The material deposited may, for example, be
`a carpet cleaning composition,
`a hard surface cleaning
`composition, or one of
`a number of compositions applied
`simultaneously, or successively, and may include a marker,
`the presence of which can be detected to provide detection
`of
`the extent of
`treatment material deposition.
`Such a
`marker may have a limited detection life,
`for example, 12,
`24 or 48 hours.
`
`Non-visible treatment may also be provided by the
`robot of
`the invention,
`for example,
`for odour control,
`antibacterial action of dust mite control.
`
`The
`
`robot
`
`preferably
`
`comprises
`
`a
`
`plurality
`
`of
`
`navigation sensors providing signals for enabling the robot
`
`to navigate over
`
`the surface,
`
`and one or more detectors
`
`the material
`the presence of
`adapted to detect
`surface
`and provide
`signals
`indicative
`thereof.
`
`on the
`The
`
`Silver Star Exhibit 1006 - 5
`
`Silver Star Exhibit 1006 - 5
`
`

`

`WO 00/04430
`
`PCT/US99/16078
`
`navigation sensors may or more_collisioninclude one
`
`
`
`sensors and/or proximity sensors.
`The collision sensors
`may
`include
`one
`or more
`lateral
`displacement
`sensors
`arranged on
`a peripheral
`sensor
`ring to provide
`360%
`collision
`detection,
`and/or
`one
`or
`more
`vertical
`displacement sensors.
`Utilising a generally circular shape together with a
`
`control regime which scans for the best direction of escape
`after
`the robot has become
`stuck (say in a corner)
`is
`especially
`advantageous.
`Furthermore,
`it
`may
`be
`additionally advantageous
`to detect
`the
`angle of
`any
`collision,
`in order to optimise the robots subsequent angle
`
`of movement away from the obstacle.
`and
`left
`The
`traction mechanism preferably includes
`right, coaxially disposed drive wheels with corresponding
`drive motors which are preferably provided with pulse-width
`
`modulated drive signals.
`an array of
`For depositing material on the surface,
`delivery ports, e.g.
`spray nozzles, may extend generally
`parallel with the drive wheel axis, preferably extending to
`
`the same lateral extent as the deposition detectors.
`
`sensors
`or more’
`one
`comprise
`The detectors may
`a section of previously
`arranged to detect
`the edge of
`deposited product.
`Suitable deposition detectors include
`one or more radiation sources and/or detectors, moisture
`
`detectors,
`reflectivity meters,
`conductivity meters etc.
`Detectors may be disposed laterally of
`the drive wheels,
`
`preferably forward thereof.
`
`The
`
`robot
`
`further preferably comprises
`
`a
`
`control
`
`system for controlling deposition of the material dependent
`on the signals received from the one or more detectors and
`sensors.
`In preferred embodiments,
`the
`control
`system
`
`functions to control deposition of
`
`the material
`
`(e.g.
`
`to
`
`by a combination of
`avoid or minimise over-application)
`the
`robot
`around
`strategies
`comprising
`a)navigating
`previously-treated areas of the surface (referred to herein
`
`Silver Star Exhibit 1006 - 6
`
`Silver Star Exhibit 1006 - 6
`
`

`

`WO 00/04430
`
`PCT/US99/16078
`
`controlling the
`b)
`and
`“navigation strategy';
`the
`as
`the deposit of
`reduce
`depositing mechanism to stop or
`fluent material on to the surface as the robot passes over
`such previously-treated areas
`(referred to herein as
`the
`“deposition rate control strategy').
`In practice,
`the
`the
`control
`system arbitrates
`between
`two
`strategies
`depending on
`the
`signals
`received from the navigation
`sensors
`and deposition detectors.
`The ability of
`the
`control system to arbitrate between the two strategies, for
`example to make
`a
`rapid judgment on whether
`to cross or
`navigate around previously-treated areas
`and whether
`to
`maintain,
`reduce or
`stop deposition accordingly,
`is an
`
`important feature for ensuring controlled deposition in the
`context of a fully autonomous robot designed to operate in
`the cluttered,
`unstructured and
`track-free
`environment
`typically found in domestic and institutional situations.
`Alternatively,
`the control system can be designed to
`control deposition purely following a deposition rate
`control
`strategy,
`in other words,
`by
`controlling the
`depositing mechanism to stop or
`reduce
`the deposit of
`fluent material on to the surface as the robot passes over
`previously-treated areas.
`Of
`course,
`systems depending
`purely on deposition rate control require less complicated
`electronics than the preferred combined-strategy systems
`
`single strategy
`the other hand,
`On
`described above.
`systems can be less efficient in terms of the time required
`to complete the task in hand.
`Preferably,
`the control
`architecture
`and
`includes
`
`a hierarchical
`system has
`one
`or more microprocessor
`
`for controlling higher-
`controllers or microcontrollers
`level
`functions,
`and providing higher-level
`instructions
`and a plurality of lower-level function modules adapted to
`receive signals
`from the sensors
`and detectors
`and to
`provide control signals in response thereto.
`The traction
`mechanism and
`product
`dispensing
`control
`signals
`are
`preferably issued to a traction mechanism controller and to
`
`Silver Star Exhibit 1006 - 7
`
`Silver Star Exhibit 1006 - 7
`
`

`

`WO 00/04430
`
`PCT/US99/16078
`
`a product dispensing controller via a manifold or bus
`arranged to receive signal
`inputs from the microprocessor
`and a plurality of sub-processors each corresponding to a
`respective navigation sensor or the like.
`By this means, a
`distributed processing system can be employed to provide a
`high level of
`flexibility in control
`strategy, whilst
`allowing simple connection of
`the sub-processors,
`thus to
`reduce the complexity and expense of
`the control
`system.
`
`The various processors preferably include neural network
`functionality
`to
`provide
`behavioural
`characteristics
`
`the
`robot,
`the
`of
`task
`chosen
`the
`to
`appropriate
`the processors preferably
`behavioural characteristics of
`being moderated by a group of generic moderators providing
`necessary arbitration between the control
`instructions from
`the
`various
`processors.
`The
`higher-level
`functions
`
`selected from
`functions
`preferably include one or more
`
`
`
`
`
`determination being~§stuck,of the robot room size
`estimation,
`clutter
`level
`determination,
`and
`battery
`
`The lower-level modules are preferably analog
`monitoring.
`neural networks which provide, for example, edge follow and
`dispense
`control
`functions,
`together,
`preferably, with
`cliff
`sensing, collision detection,
`speed reduction and
`random movement functions.
`
`One example of a self-propelled robot constructed in
`accordance with the present
`invention,
`and its method of
`operation, will
`now be described with reference to the
`accompanying drawings in which:-
`Figure 1 is an underneath plan view of the robot;
`Figure 2 is a functional diagram of the robot; and
`Figures 3A-C illustrate neural net aspects of part of
`
`the robot's control system.
`
`the robot of the present
`As can be seen from Figure 1,
`example is substantially circular in overall plan view.
`A
`simple plate-like chassis 1
`supports both the mechanical
`and electrical components of
`the robot.
`The plate-like
`chassis 1 supports the body 2 of
`the robot on resilient
`
`Silver Star Exhibit 1006 - 8
`
`Silver Star Exhibit 1006 - 8
`
`

`

`WO 00/04430
`
`PCT/US99/16078
`
`rubber mountings 3 which allow the body to move relative to
`the chassis when a force is applied,
`eg by collision with
`an object,
`to a sensor ring 20 which is disposed around the
`periphery of the body.
`Four displacement sensors 4 placed
`at
`90% intervals
`around
`the
`robot measure
`lateral
`
`the body 2 relative to the chassis 1 and
`displacement of
`inform the control
`system of contact with an external
`object.
`The displacement sensors 4 are based on linear
`Hall Effect devices which produce
`a voltage which
`is
`proportional to the strength of the magnetic field in which
`they immersed.
`Each sensor consists of a small permanent
`magnet mounted on the body shell support ring 20 and a Hall
`Effect device mounted on the main chassis 1. When the body
`moves with respect
`to the chassis
`(as happens during a
`collision)
`the voltage produced by the Hall Effect device
`varies and can be used to signal the control system that an
`object has been encountered.
`By examining the signals from
`all four sensors the angle and magnitude of
`the collision
`can be deduced.
`These sensors allow displacements in the
`order of 0.1 mm to be reliably detected.
`A fifth sensor
`
`same
`the
`of
`18,
`measures vertical
`accommodate
`forces
`
`the displacement
`as
`type
`displacement
`of
`the
`body
`produced
`by
`objects which
`
`sensors
`shell
`are
`
`4,
`to
`of
`
`insufficient height to cause lateral body movement.
`
`In an
`
`alternative construction,
`
`these sensors may be superseded
`
`by a single custom-built sensor which can measure lateral
`and
`vertical
`displacement
`simultaneously.
`Such
`an
`integrated sensor may be optical
`in nature utilising an
`array of photo detectors mounted on the chassis and a light
`source which is mounted on the body support ring.
`A
`single
`forward
`facing time-of-flight ultrasound
`sensor 13 is mounted at the front of the robot and is used
`
`to allow the robot to gather more information regarding its
`surroundings
`than can
`be
`achieved by
`the displacement
`sensors 4 alone.
`This ultrasound sensor 13 is based on a
`
`Polaroid® ranging module Polaroid 6500 series sonar ranging
`
`Silver Star Exhibit 1006 - 9
`
`Silver Star Exhibit 1006 - 9
`
`

`

`WO00/04430
`
`PCT/US99/16078
`
`the data from which is
`device, Polaroid reference 615077,
`pre-processed by a dedicated unit 5 on which the sensor 13
`is located.
`An ultrasonic sensor unit 5, containing the
`
`ultrasonic sensor
`
`13
`
`itself
`
`and
`
`a
`
`suitable electronic
`
`interface, are mounted on the body to provide proximity
`information to the robot's control system.
`Left
`and right motors 6,
`7 are provided to drive
`corresponding left and right wheels 8,
`9 each with a soft
`rubber tyre, via an integral reduction gearbox,
`to provide
`motive power to the robot.
`A single castor 10 mounted at
`the rear of the robot completes the drive/movement system
`
`and allows the chassis to move
`
`forwards or backwards and
`
`the
`rotate on the spot. Varying the rotational speed of
`left and right motors 6,
`7 allows the robot
`to be steered
`in any direction.
`The speed of the motors is controlled by
`pulse width modulating the voltages applied to the motors.
`This involves switching the motor current on and off very
`rapidly (100,000 times a second)
`and varying the ratio of
`'on'
`time to ‘off'
`time.
`This is a very efficient way to
`control the power to the motors and hence their speed.
`Power for the robot,
`including the motors 6, 7 and the
`
`control system is provided by means of a battery pack 11
`mounted on the chassis 1.
`To protect the components of the
`robot
`from tampering and from damage a cover or housing
`(not
`shown)
`is attached to the body 2
`to house the robot
`components.
`In the preferred embodiment,
`this is part-
`spherical or dome-like in shape.
`A row of spray nozzles 16 and a pump 115 (not shown in
`Figure 1) provide a means of dispensing treating fluid on
`
`to the surface to be treated and detectors 14,15,17 are
`provided to detect the presence of the treating fluid (or a
`suitable additional marker fluid).
`The three sensor units
`14, 15, 17, one placed in front of each of the drive wheels
`and
`the
`third 17
`placed centrally,
`emit
`light at
`a
`wavelength which excites a fluorescent dye in the product
`being detected.
`These sensor units incorporate a pair of
`
`Silver Star Exhibit 1006 - 10
`
`Silver Star Exhibit 1006 - 10
`
`

`

`WO 00/04430
`
`PCT/US99/16078
`
`light sensitive devices positioned at 90® to the robot's
`direction of travel and spaced 20mm apart, which can detect
`light produced by the fluorescent dye.
`By examining the
`intensity of
`the light detected by these devices the edge
`of
`a
`section of previously deposited product
`can
`be
`
`detected
`
`and
`
`hence
`
`followed.
`
`In
`
`an
`
`alternative
`
`construction,
`
`the three sensor units 14,
`
`15,
`
`17 pass
`
`a
`
`small electrical current
`
`through the floor
`
`covering by
`
`virtue of an array of stainless steel contacts which are
`designed to glide over
`the floor covering surface.
`The
`conductivity of the floor covering will vary depending upon
`
`whether or not it has recently been sprayed with product.
`
`the
`the floor covering,
`By examining the conductivity of
`edge of previously deposited product can be detected and
`hence followed.
`
`In an alternative construction,
`
`in which fluid is to
`
`the positioning of the
`be dispensed to an edge or corner,
`sprays
`is modified.
`The modification is such that
`the
`spray is able to dispense to the edge of
`the robot or
`
`for example, either by positioning nozzles at the
`beyond,
`very periphery of
`the underside or by additional nozzles
`which protrude from the casing and are directed such that
`they spray beyond the perimeter of the robot.
`The robot's control
`system comprises various circuit
`boards and components which are not
`shown in Figure 1
`in
`detail,
`but which
`are broadly
`indicated by
`reference
`
`numerals 12 in Figure 1.
`
`The control
`
`system will now be described in further
`
`detail.
`
`Two purposes of
`
`the control
`
`system of an autonomous
`
`mobile robot such as that of the example are to allow the
`
`robot
`
`to move within a physical environment
`
`in safety and
`
`To do this the robot
`to enable it to perform useful tasks.
`must be aware of its immediate surroundings and be able to
`react
`to particular circumstances in particular ways.
`A
`robot
`intended for
`an unconstrained domestic environment
`
`Silver Star Exhibit 1006 - 11
`
`Silver Star Exhibit 1006 - 11
`
`

`

`WO 00/04430
`
`PCT/US99/16078
`
`10
`
`such as a collision
`to have certain basic skills,
`needs
`it
`to stop
`upon
`detection skill, which might
`cause
`collision with an object
`and then take evasive action
`before resuming its previous activity.
`
`the sensors 4, 18,
`In the case of collision detection,
`13, which sense impacts with and proximity to objects, will
`inform the control system of
`the angle of
`impact and its
`force.
`The control system must react very quickly to this
`stimulus and prevent any further motion in this direction.
`A conventional approach to this problem would be to have a
`computer monitor
`the collision sensors
`and act upon the
`data to stop the motors
`and then perform some
`form of
`avoidance manoeuvre.
`This
`is perfectly feasible, but
`if
`the same computer
`is required simultaneously to perform
`other
`tasks,
`for example,
`such as
`in the present case,
`monitoring
`other
`sensors
`and
`performing
`navigational
`mathematics,
`it soon reaches a point where the speed and
`power
`of
`the
`on-board
`computer
`required
`becomes
`prohibitively expensive
`if
`reaction times
`are
`to be
`
`acceptable.
`is
`invention,
`The alternative, adopted in the present
`to use discrete modules
`that perform functions in a way
`analogous to the reflexes of a biological organism.
`The
`advantage of
`this system are obvious:
`the main processor
`can merely issue high level commands such as move or turn
`and is left free to perform other abstract tasks.
`This alternative is a form of hierarchical distributed
`
`processing and allows the control system to be composed of
`simple modules
`that
`together yield faster response times
`than a non-distributed system of
`the same cost.
`Another
`significant
`advantage of distributed processing is
`its
`inherent robustness.
`If a system employing a conventional
`single processor approach suffers a failure,
`it can leave
`the system in an unsafe state, which in the case of a robot
`might allow it
`to crash into objects or people.
`The
`distributed approach can be designed so as to have a much
`
`Silver Star Exhibit 1006 - 12
`
`Silver Star Exhibit 1006 - 12
`
`

`

`WO 00/04430
`
`PCT/US99/16078
`
`el
`
`rendering the occurrence
`greater degree of fault tolerance,
`of complete system failures much less likely.
`Distributed
`processing
`can
`be
`implemented
`
`using
`
`conventional computers connected together by some
`
`form of
`
`these tend to be expensive to design and
`network, but
`The approach adopted in the present
`invention
`implement.
`is to simulate biological neural networks in real analogue
`
`hardware to provide a system that consists of behavioural
`
`modules, which are designed to perform individual
`tasks.
`These behaviours are managed by a simple micro controller,
`which performs higher
`level
`tasks
`such as mathematical
`functions to estimate room size or a strategy for escaping
`
`from under a table.
`
`now be described with
`system 100 will
`The control
`reference to Figures 2 and 3.
`Figure 2
`illustrates the
`
`functional relationship of the control system components.
`
`The control behaviours used on
`
`the
`
`robot
`
`can be
`
`divided into two basic types,
`
`Low Level
`
`and High Level.
`
`Low Level
`
`behaviours
`
`are
`
`implemented
`
`in
`
`hardware
`
`as
`
`discrete neural blocks or modules 101-105, while High Level
`
`behaviours
`
`are
`
`software
`
`algorithms
`
`running on
`
`a micro
`
`controller 106.
`
`The functions of the Low level behaviour modules 101-
`
`105 are now described in detail:-
`
`Cliff - To prevent the robot falling down stairs it is
`
`equipped with four cliff detectors 21 which warn of
`vertical hazards
`and provide signals
`to the cliff
`
`behaviour module 101.
`
`The cliff detectors
`
`21 are
`
`active infra red proximity sensors which comprise a
`
`modulated light source which emits a beam of infra red
`light directed at the target (in this case the floor),
`and an infra red detector which monitors the intensity
`
`of the light which is reflected. When the sensor is
`
`directed over a cliff the intensity of
`
`the reflected
`
`informs
`and the sensor
`light decreases
`system of
`the hazard.
`This behavioural
`
`the control
`function has
`
`Silver Star Exhibit 1006 - 13
`
`Silver Star Exhibit 1006 - 13
`
`

`

`WO 00/04430
`
`PCT/US99/16078
`
`12
`
`to
`operates
`active
`high priority and when
`very
`manoeuvre the robot away from the hazard and return it
`to a course which is modified to avoid cliff type
`
`drops.
`Edge Follow - The Edge Follow module 104 provides a
`behavioural
`function which uses
`information from the
`
`sensors 14,15,17 which allow the robot
`
`to find the
`
`edge of a previously treated area (as described above)
`
`and to travel along that edge to produce a faster scan
`
`of the floor surface.
`
`Random - In the absence of any edges the robot moves
`
`in a random direction under
`
`the action of
`
`a
`
`random
`
`movement module 114 until an object is encountered or
`the edge follow behaviour is activated.
`Collide - The collision detection module
`
`takes
`
`102
`
`from the displacement sensors 4,18 and operates
`input
`so that upon encountering an obstacle the robot stops,
`reverses a small distance,
`then turns away from the
`object
`in a direction that depends upon the angle of
`impact, which is determined from the signals of
`the
`displacement sensors 4,18.
`is detected by the
`Reduce Speed - When
`an object
`ultrasound sensor unit 5 within a pre-set range limit,
`the forward speed of
`the robot
`is reduced by the
`
`Reduce Speed module 103 to minimise the impact force
`generated when contact with the object occurs.
`inputs
`Dispense - A dispense control module 105 has
`from a fluid level sensor 203 and sensors 14, 15,
`17
`
`via the Edge Follow module 104.
`15,
`17
`report untreated carpet
`travel
`the
`treatment
`chemical
`
`If the UV sensors 14,
`in the direction of
`is
`dispensed until
`
`treated areas are encountered or fluid level reaches a
`
`lower limit.
`
`the
`determined within
`are
`behaviours
`level
`High
`microcontroller 106 and comprise the following functional
`
`modules:-
`
`Silver Star Exhibit 1006 - 14
`
`Silver Star Exhibit 1006 - 14
`
`

`

`WO 00/04430
`
`PCT/US99/16078
`
`13
`
`Stuck - A routine 107 determines if there have been
`
`more than a chosen number of collisions in a select
`
`period and causes
`
`the robot
`
`to stop and use
`
`the
`
`ultrasound range finder 5,
`13
`to find the longest
`clear path and move in that direction.
`The robot will
`rotate on the spot, by operating the wheels 8,
`9
`in
`opposite directions,
`looking for
`the longest clear
`path. When the best direction is discovered the robot
`will move off in that direction.
`
`Estimate Room size - By using statistics gathered from
`the ultrasound sensor
`13
`and measuring
`the
`time
`
`between collisions the routine 108 is able to estimate
`
`the area of the room.
`
`This is used to determine how
`
`long the robot should take to treat a particular room.
`Estimate clutter
`level
`- By comparing estimates of
`
`room size against collisions per minute a routine 109
`is able to deduce a factor describing the complexity
`of the room.
`This can then be used to modify the run
`
`time to allow for the level of clutter.
`
`Battery Monitor - A battery monitor routine 110 checks
`
`the battery by monitoring the output
`the state of
`voltage and current.
`It uses
`this information to
`estimate how long the battery will be able to support
`
`the robot's systems before a
`
`re-charge is needed.
`
`When
`
`the monitor
`
`routine decides
`
`that
`
`the battery
`
`reliable
`where’
`point
`the
`approaching
`is
`state
`operation is no longer possible,
`the user is warned by
`illumination of a battery low indicator.
`If the robot
`is
`allowed
`to continue
`to operate without
`being
`re-charged the monitor
`routine will
`shut
`the robot
`
`down
`
`in a
`
`safe and controlled fashion when
`
`power
`
`levels reach a predetermined point. Nickel Cadmium or
`
`Nickel Metal Hydride
`
`batteries
`
`require
`
`careful
`
`charging to ensure maximum capacity and life span and
`
`the monitor routine also controls the charging cycle
`
`of the battery to ensure that these needs are met.
`
`Silver Star Exhibit 1006 - 15
`
`Silver Star Exhibit 1006 - 15
`
`

`

`WO 00/04430
`
`PCT/US99/16078
`
`14
`
`Traditionally neural network designers have insisted
`that every neuron in a network is connected to every other
`neuron in that network. Whilst this allows the network the
`
`level of flexibility, very many (even as high as
`greatest
`90%) of these connections will never be used.
`The present
`system allows
`pre-configured
`neural
`networks
`to
`be
`connected together in a much less complex way allowing the
`
`behaviour of
`
`the
`
`robot
`
`to dynamically
`
`adjust
`
`to the
`
`immediate environment in a continuous fashion.
`
`an
`comprises
`so-called "Manifold Architecture"
`This
`analogue bus or manifold 111, connecting all the behaviour
`modules
`101-105
`and their associated actuators
`to each
`
`Four generic moderators arbitrate between the
`other.
`behaviours, and give rise to a prototype behaviour of their
`own which regulates the overall activity of the robot via a
`motor controller 112 and dispensing fluid pump controller
`
`These generic moderators sum all
`113 driving the pump 115.
`the excitatory and inhibitory inputs and apply a non-linear
`transfer function to the results.
`The outputs from these
`
`moderators form the inputs to the motor controllers.
`
`In order
`
`to explain the function of
`
`the manifold
`
`architecture, it is necessary to describe the basic neural
`aspects of
`the control
`system.
`Figures
`3A-C will
`be
`referenced for this purpose.
`
`3A) has three basic types of
`A single neuron (see Fig.
`connections, excitatory inputs which cause the neuron to
`'fire',
`inhibitory inputs which suppress activity and the
`output which
`represents
`the
`state
`of
`the
`neuron.
`
`such as
`Additionally neurons may have other properties
`Decay which causes the output to fall slowly over time, and
`Threshold which suppresses all output until the sum of all
`the input exceeds a certain level.
`simplified
`a
`example)
`Figure
`3B shows
`(by way of
`representation of
`the collide behaviour and the manifold
`system in neural notation.
`
`Silver Star Exhibit 1006 - 16
`
`Silver Star Exhibit 1006 - 16
`
`

`

`WO 00/04430
`
`PCT/US99/16078
`
`15
`
`3B as
`The collision sensors 4 are represented in Fig.
`1, 2,
`3 and 4 and are buffered and normalised by sensor
`
`pre-processors 5, 6,
`7 and 8.
`The outputs of the sensor
`pre-processors are each fed into a single neuron 9, 10, 11
`and 12 configured as a pulse stretcher with a time constant
`of approximately 5 seconds.
`The outputs of these neurons
`
`are connected to the rest of the network formed by neurons
`
`and transfer
`to 28 where the pattern of connections,
`13
`characteristics of
`the neurons g