InvertMotorBigModel Class Reference

class for robot controller is based on InvertMotorNStep More...

#include <invertmotorbigmodel.h>

Inheritance diagram for InvertMotorBigModel:

[legend]Collaboration diagram for InvertMotorBigModel:


Public Member Functions
	InvertMotorBigModel (const InvertMotorBigModelConf &conf=getDefaultConf())
virtual void	init (int sensornumber, int motornumber, RandGen *randGen=0)
	initialisation of the controller with the given sensor/ motornumber Must be called before use.
virtual	~InvertMotorBigModel ()
virtual int	getSensorNumber () const
	returns the number of sensors the controller was initialised with or 0 if not initialised
virtual int	getMotorNumber () const
	returns the mumber of motors the controller was initialised with or 0 if not initialised
virtual void	step (const sensor , int number_sensors, motor , int number_motors)
	performs one step (includes learning).
virtual void	stepNoLearning (const sensor , int number_sensors, motor , int number_motors)
	performs one step without learning. Calulates motor commands from sensor inputs.
virtual bool	store (FILE *f) const
	stores the controller values to a given file.
virtual bool	restore (FILE *f)
	loads the controller values from a given file.
virtual iparamkeylist	getInternalParamNames () const
	The list of the names of all internal parameters given by getInternalParams().
virtual iparamvallist	getInternalParams () const
virtual ilayerlist	getStructuralLayers () const
	Specifies which parameter vector forms a structural layer (in terms of a neural network) The ordering is important.
virtual iconnectionlist	getStructuralConnections () const
	Specifies which parameter matrix forms a connection between layers (in terms of a neural network) The orderning is not important.
virtual paramval	getParam (const paramkey &key) const
virtual bool	setParam (const paramkey &key, paramval val)
virtual paramlist	getParamList () const
	The list of all parameters with there value as allocated lists.
virtual void	setMotorTeachingSignal (const motor *teaching, int len)
	The given motor teaching signal is used for this timestep.
virtual void	setSensorTeachingSignal (const sensor *teaching, int len)
	The given sensor teaching signal (distal learning) is used for this timestep.
void	getLastMotors (motor *motors, int len)
void	kwtaInhibition (matrix::Matrix &weightmatrix, unsigned int k, double damping)
	k-winner take all inhibition for synapses.
void	limitC (matrix::Matrix &weightmatrix, unsigned int rfSize)
	sets all connections to zero which are further away then rfSize.
Static Public Member Functions
static InvertMotorBigModelConf	getDefaultConf ()
Protected Member Functions
virtual void	fillBuffersAndControl (const sensor x_, int number_sensors, motor y_, int number_motors)
	puts the sensors in the ringbuffer, generate controller values and put them in the
virtual void	calcEtaAndBufferIt (int delay)
	calculates the first shift into the motor space useing delayed motor values.
virtual void	learnController ()
	learn H,C with motors y and corresponding sensors x
virtual void	calcCandHUpdates (matrix::Matrix &C_update, matrix::Matrix &H_update, int y_delay)
	calculates the Update for C and H
virtual void	updateCandH (const matrix::Matrix &C_update, const matrix::Matrix &H_update, double squashSize)
	updates the matrix C and H
virtual void	learnModel (int delay)
	learn A, (and S) using motors y and corresponding sensors x
virtual void	management ()
	handles inhibition damping etc.
virtual matrix::Matrix	calculateControllerValues (const matrix::Matrix &x_smooth)
	returns controller output for given sensor values
matrix::Matrix	calcDerivatives (const matrix::Matrix *buffer, int delay)
	Calculates first and second derivative and returns both in on matrix (above).
Protected Attributes
unsigned short	number_sensors
unsigned short	number_motors
matrix::Matrix	A
	Model Matrix (motors to sensors).
matrix::Matrix	S
	additional Model Matrix (sensors derivatives to sensors)
matrix::Matrix	C
	Controller Matrix.
matrix::Matrix	H
	Controller Bias.
NoiseGenerator *	BNoiseGen
	Noisegenerator for noisy bias.
matrix::Matrix	R
	C*A.
matrix::Matrix	SmallID
	small identity matrix in the dimension of R
matrix::Matrix	xsi
	current output error
double	xsi_norm
	norm of matrix
double	xsi_norm_avg
	average norm of xsi (used to define whether Modell learns)
double	pain
	if the modelling error (xsi) is too high we have a pain signal
matrix::Matrix *	x_buffer
matrix::Matrix *	y_buffer
matrix::Matrix *	eta_buffer
matrix::Matrix	zero_eta
matrix::Matrix	x_smooth
matrix::Matrix	y_teaching
	teaching motor signal
bool	useTeaching
	flag whether there is an actual teachning signal or not
int	t_rand
	initial random time to avoid syncronous management of all controllers
int	managementInterval
	interval between subsequent management function calls
paramval	inhibition
	inhibition strength for sparce kwta strategy (is scaled with epsC)
paramval	kwta
	(int) number of synapses that get strengthend
paramval	limitRF
	(int) receptive field of motor neurons (number of offcenter sensors) if null then no limitation. Mutual exclusive with inhibition
paramval	dampS
	damping of S matrix
InvertMotorBigModelConf	conf

Detailed Description

class for robot controller is based on InvertMotorNStep

direct inversion

motor space

multilayer,nonlinear model

Constructor & Destructor Documentation

InvertMotorBigModel ( const InvertMotorBigModelConf & conf = getDefaultConf() )

~InvertMotorBigModel ( ) [virtual]

Member Function Documentation

void calcCandHUpdates	(	matrix::Matrix &	C_update,
		matrix::Matrix &	H_update,
		int	y_delay
	)			`[protected, virtual]`

calculates the Update for C and H

Matrix calcDerivatives	(	const matrix::Matrix *	buffer,
		int	delay
	)			`[protected]`

Calculates first and second derivative and returns both in on matrix (above).

We use simple discrete approximations:

$f'(x) = (f(x) - f(x-1)) / 2$

$f''(x) = f(x) - 2f(x-1) + f(x-2)$

where we have to go into the past because we do not have f(x+1). The scaling can be neglegted.

void calcEtaAndBufferIt ( int delay ) [protected, virtual]

calculates the first shift into the motor space useing delayed motor values.

Matrix calculateControllerValues ( const matrix::Matrix & x_smooth ) [protected, virtual]

returns controller output for given sensor values

Parameters:

x_smooth

smoothed sensors Matrix(number_channels,1)

void fillBuffersAndControl	(	const sensor *	x_,
		int	number_sensors,
		motor *	y_,
		int	number_motors
	)			`[protected, virtual]`

puts the sensors in the ringbuffer, generate controller values and put them in the

static InvertMotorBigModelConf getDefaultConf ( ) [inline, static]

list< Inspectable::iparamkey > getInternalParamNames ( ) const [virtual]

The list of the names of all internal parameters given by getInternalParams().

The naming convention is "v[i]" for vectors and "A[i][j]" for matrices, where i, j start at 0.

Returns:: : list of keys

Reimplemented from Inspectable.

list< Inspectable::iparamval > getInternalParams ( ) const [virtual]

Returns:: : list of values

Reimplemented from Inspectable.

void getLastMotors	(	motor *	motors,
		int	len
	)

virtual int getMotorNumber ( ) const [inline, virtual]

returns the mumber of motors the controller was initialised with or 0 if not initialised

Implements AbstractController.

Configurable::paramval getParam ( const paramkey & key ) const [virtual]

Configurable::paramlist getParamList ( ) const [virtual]

The list of all parameters with there value as allocated lists.

Note that these are only parameters that are managed manually (with setParam, getParam)

See also:: getAllParamNames()

Returns:: list of key-value pairs

Reimplemented from Configurable.

virtual int getSensorNumber ( ) const [inline, virtual]

returns the number of sensors the controller was initialised with or 0 if not initialised

Implements AbstractController.

list< Inspectable::IConnection > getStructuralConnections ( ) const [virtual]

Specifies which parameter matrix forms a connection between layers (in terms of a neural network) The orderning is not important.

Returns:: : list of layer names with dimension

Reimplemented from Inspectable.

list< Inspectable::ILayer > getStructuralLayers ( ) const [virtual]

Specifies which parameter vector forms a structural layer (in terms of a neural network) The ordering is important.

The first entry is the input layer and so on.

Returns:: : list of layer names with dimension

Reimplemented from Inspectable.

void init	(	int	sensornumber,
		int	motornumber,
		RandGen *	randGen = `0`
	)			`[virtual]`

initialisation of the controller with the given sensor/ motornumber Must be called before use.

The random generator is optional.

Implements AbstractController.

void kwtaInhibition	(	matrix::Matrix &	weightmatrix,
		unsigned int	k,
		double	damping
	)

k-winner take all inhibition for synapses.

k largest synapses are strengthed and the rest are inhibited. strong synapes are scaled by 1+(damping/k) and weak synapses are scaled by 1-(damping/(n-k)) where n is the number of synapes

Parameters:

	weightmatrix	reference to weight matrix. Synapses for a neuron are in one row. The inhibition is done for all rows independently
	k	number of synapes to strengthen
	damping	strength of supression and exitation (typically 0.001)

void learnController ( ) [protected, virtual]

learn H,C with motors y and corresponding sensors x

void learnModel ( int delay ) [protected, virtual]

learn A, (and S) using motors y and corresponding sensors x

void limitC	(	matrix::Matrix &	weightmatrix,
		unsigned int	rfSize
	)

sets all connections to zero which are further away then rfSize.

If rfSize == 1 then only main diagonal is left. If rfSize = 2: main diagonal and upper and lower side diagonal are kept and so on and so forth.

void management ( ) [protected, virtual]

handles inhibition damping etc.

bool restore ( FILE * f ) [virtual]

loads the controller values from a given file.

Implements Storeable.

void setMotorTeachingSignal	(	const motor *	teaching,
		int	len
	)			`[virtual]`

The given motor teaching signal is used for this timestep.

It is used as a feed forward teaching signal for the controller. Please note, that the teaching signal has to be given each timestep for a continuous teaching process.

bool setParam	(	const paramkey &	key,
		paramval	val
	)			`[virtual]`

void setSensorTeachingSignal	(	const sensor *	teaching,
		int	len
	)			`[virtual]`

The given sensor teaching signal (distal learning) is used for this timestep.

First the belonging motor teachung signal is calculated by the inverse model. See setMotorTeachingSignal

void step	(	const sensor *	,
		int	number_sensors,
		motor *	,
		int	number_motors
	)			`[virtual]`

performs one step (includes learning).

Calulates motor commands from sensor inputs.

Implements AbstractController.

void stepNoLearning	(	const sensor *	,
		int	number_sensors,
		motor *	,
		int	number_motors
	)			`[virtual]`

performs one step without learning. Calulates motor commands from sensor inputs.

Implements AbstractController.

bool store ( FILE * f ) const [virtual]

stores the controller values to a given file.

Implements Storeable.

void updateCandH	(	const matrix::Matrix &	C_update,
		const matrix::Matrix &	H_update,
		double	squashSize
	)			`[protected, virtual]`

updates the matrix C and H

Member Data Documentation

matrix::Matrix A [protected]

Model Matrix (motors to sensors).

NoiseGenerator* BNoiseGen [protected]

Noisegenerator for noisy bias.

matrix::Matrix C [protected]

Controller Matrix.

InvertMotorBigModelConf conf [protected]

paramval dampS [protected]

damping of S matrix

matrix::Matrix* eta_buffer [protected]

matrix::Matrix H [protected]

Controller Bias.

paramval inhibition [protected]

inhibition strength for sparce kwta strategy (is scaled with epsC)

paramval kwta [protected]

(int) number of synapses that get strengthend

paramval limitRF [protected]

(int) receptive field of motor neurons (number of offcenter sensors) if null then no limitation. Mutual exclusive with inhibition

int managementInterval [protected]

interval between subsequent management function calls

unsigned short number_motors [protected]

unsigned short number_sensors [protected]

double pain [protected]

if the modelling error (xsi) is too high we have a pain signal

matrix::Matrix R [protected]

C*A.

matrix::Matrix S [protected]

additional Model Matrix (sensors derivatives to sensors)

matrix::Matrix SmallID [protected]

small identity matrix in the dimension of R

int t_rand [protected]

initial random time to avoid syncronous management of all controllers

bool useTeaching [protected]

flag whether there is an actual teachning signal or not

matrix::Matrix* x_buffer [protected]

matrix::Matrix x_smooth [protected]

matrix::Matrix xsi [protected]

current output error

double xsi_norm [protected]

norm of matrix

double xsi_norm_avg [protected]

average norm of xsi (used to define whether Modell learns)

matrix::Matrix* y_buffer [protected]

matrix::Matrix y_teaching [protected]

teaching motor signal

matrix::Matrix zero_eta [protected]

The documentation for this class was generated from the following files:

Generated on Fri Oct 30 16:29:02 2009 for Robot Simulator of the Robotics Group for Self-Organization of Control by

1.4.7

InvertMotorBigModel Class Reference

Public Member Functions

Static Public Member Functions

Protected Member Functions

Protected Attributes

Detailed Description

Constructor & Destructor Documentation

Member Function Documentation

Member Data Documentation