Apparatus incorporating recursive estimators

Claim 1

An apparatus for measuring the state of a dynamic system comprising: (a) sensing means for producing a sequence of measurements or sets of measurements which are dependent upon the current state of the said system in the presence of noise and other sources of error being sources which are representable by a statistical model; (b) first computing means for determining a log-likelihood function associated with each of the said measurements or sets of measurements in accordance with the said statistical model; (c) second computing means for obtaining the sum of said log-likelihood function and a residual function which is initially set to be identically zero; (d) third computing means using a recursive estimator to model the dynamic behaviour of the system; (e) fourth computing means for determining a quadratic approximation to the said sum of the log-likelihood function and the residual function, the fit being weighted in dependence upon the present uncertainty in the state of the dynamic system as evidenced by the output of the third computing means; (f) fifth computing means for deriving from the said quadratic approximation parameters to serve as the basis for a measurement update within the third computing means; (g) sixth computing means for subtracting the said quadratic approximation from the said sum of the log-likelihood function and the residual function which it approximates so as to form a new residual function; (h) seventh computing means for transforming the said residual function to take into account dynamic change in the state of the system during the interval between successive measurements or sets of measurements in the said measurement sequence; and (i) eighth computing means for determining whether the said quadratic approximation produced by said fourth computing means is suitable to serve as an input to said third computing means and being operative, if it is determined that the quadratic approximation is suitable for the aforesaid purpose, to activate the fifth and sixth computing means and otherwise to use as an input to the seventh computing means the sum of the log-likelihood function and the residual function, which sum is used as input to the fourth computing means and is taken as the new said residual function.

Claim 2

An apparatus for measuring the state of a dynamic system comprising: (a) sensing means for producing a sequence of measurements or sets of measurements which are dependent upon the current state of the said system in the presence of noise and other sources of error being sources which are representable by a statistical model; (b) first computing means for determining a log-likelihood function associated with each of the said measurements or sets of measurements in accordance with the said statistical model; (c) second computing means for obtaining the sum of said log-likelihood function and a residual function which is initially set to be identically zero; (d) third computing means using a recursive estimator to model the dynamic behaviour of the system; (e) fourth computing means for determining a quadratic approximation to the said sum of the log-likelihood function and the residual function, the fit being weighted in dependence upon the present uncertainty in the state of the dynamic system as evidenced by the output of the third computing means; (f) fifth computing means for deriving from the said quadratic approximation parameters to serve as the basis for a measurement update within the third computing means; (g) sixth computing means for subtracting the said quadratic approximation from the said sum of the log-likelihood function and the residual function which it approximates so as to form a new residual function; (h) seventh computing means for transforming the said residual function to take into account dynamic change in the state of the system during the interval between successive measurements or sets of measurements in the said measurement sequence; and (i) eighth computing means for determining whether the said quadratic approximation produced by said fourth computing means is suitable to serve as an input to said third computing means and being operative, if it is determined that the quadratic approximation is suitable for the aforesaid purpose, to activate the fifth and sixth computing means and otherwise to use as an input to the seventh computing means the sum of the log-likelihood function and the residual function, which sum is used as input to the fourth computing means and is taken as the new said residual function.

Claim 5

An apparatus for measuring the state of a dynamic system comprising: (a) sensing means for producing a sequence of measurements or sets of measurements which are dependent upon the current state of the said system in the presence of noise and other sources of error being sources which are representable by a statistical model; (b) first computing means for determining a likelihood function associated with each of the said measurements or sets of measurements in accordance with the said statistical model; (c) second computing means for obtaining the product of the said likelihood function and a residual function which is initially set to be identically unity; (d) third computing means using a recursive estimator to model the dynamic behaviour of the system; (e) fourth computing means for determining a Gaussian approximation to the said product of the likelihood function and the residual function, the fit being weighted in dependence upon the present uncertainty in the state of the dynamic system as evidenced by the output of the said third computing means; (f) fifth computing means for deriving from the said Gaussian approximation parameters to serve as the basis for a measurement update within the third computing means; (g) sixth computing means for dividing the said Gaussian approximation into the said product of the likelihood function and the residual function which it approximates so as to form a new residual function; (h) seventh computing means for transforming the said residual function to take into account dynamic change in the state of the system during the interval between successive measurements or sets of measurements in the said measurement sequence; and (i) eighth computing means for determining whether the said Gaussian approximation is suitable to serve as an input to said third computing means and being operative, if it is determined that the Gaussian approximation is suitable for the aforesaid purpose, to activate the fifth and sixth computing means and otherwise to use as an input to the seventh computing means the product of the likelihood function and the residual function, which product is used as input to the fourth computing means and is taken as the new said residual function.

Claim 6

An apparatus for measuring the state of a dynamic system comprising: (a) sensing means for producing a sequence of measurements or sets of measurements which are dependent upon the current state of the said system in the presence of noise and other sources of error being sources which are representable by a statistical model; (b) first computing means for determining a likelihood function associated with each of the said measurements or sets of measurements in accordance with the said statistical model; (c) second computing means for obtaining the product of the said likelihood function and a residual function which is initially set to be identically unity; (d) third computing means using a recursive estimator to model the dynamic behaviour of the system; (e) fourth computing means for determining a Gaussian approximation to the said product of the likelihood function and the residual function, the fit being weighted in dependence upon the present uncertainty in the state of the dynamic system as evidenced by the output of the said third computing means; (f) fifth computing means for deriving from the said Gaussian approximation parameters to serve as the basis for a measurement update within the third computing means; (g) sixth computing means for dividing the said Gaussian approximation into the said product of the likelihood function and the residual function which it approximates so as to form a new residual function; (h) seventh computing means for transforming the said residual function to take into account dynamic change in the state of the system during the interval between successive measurements or sets of measurements in the said measurement sequence; and (i) eighth computing means for determining whether the said Gaussian approximation is suitable to serve as an input to said third computing means and being operative, if it is determined that the Gaussian approximation is suitable for the aforesaid purpose, to activate the fifth and sixth computing means and otherwise to use as an input to the seventh computing means the product of the likelihood function and the residual function, which product is used as input to the fourth computing means and is taken as the new said residual function.

Claim 15

A method of measuring the state of a dynamic system comprising the steps of: (a) producing a sequence of measurements or sets of measurements which are dependent upon the current state of the said system in the presence of noise and other sources of error being sources which are representable by a statistical model; (b) determining a log-likelihood function associated with each of the said measurements or sets of measurements in accordance with the said statistical model; (c) obtaining the sum of said log-likelihood function and a residual function which is initially set to be identically zero; (d) transforming the said residual function to take into account dynamic change in the state of the system during the interval between successive measurements or sets of measurements in the said measurement sequence; (e) utilising a recursive estimator to model the dynamic behaviour of the system; (f) determining a quadratic approximation to the said sum of the log-likelihood function and the residual function, the fit being weighted in dependence upon the present uncertainty in the state of the dynamic system as evidenced by the modelled behavior of the system; (g) determining whether the said quadratic approximation is suitable to serve as an input for said modelling of the behaviour of the system, and if it is determined that the quadratic approximation is suitable for the aforesaid purpose; (h) deriving from the said quadratic approximation parameters to serve as the basis for a measurement update for the modelling of the behaviour of the system; (i) subtracting the said quadratic approximation from the sum of the log-likelihood function and the residual function which it approximates so as to form a new residual function; and (j) otherwise using said sum of the log-likelihood function and the residual function used for determining said quadratic approximation as the new said residual function.

Claim 16

A method of measuring the state of a dynamic system comprising the steps of: (a) producing a sequence of measurements or sets of measurements which are dependent upon the current state of the said system in the presence of noise and other sources of error being sources which are representable by a statistical model; (b) determining a log-likelihood function associated with each of the said measurements or sets of measurements in accordance with the said statistical model; (c) obtaining the sum of said log-likelihood function and a residual function which is initially set to be identically zero; (d) transforming the said residual function to take into account dynamic change in the state of the system during the interval between successive measurements or sets of measurements in the said measurement sequence; (e) utilising a recursive estimator to model the dynamic behaviour of the system; (f) determining a quadratic approximation to the said sum of the log-likelihood function and the residual function, the fit being weighted in dependence upon the present uncertainty in the state of the dynamic system as evidenced by the modelled behavior of the system; (g) determining whether the said quadratic approximation is suitable to serve as an input for said modelling of the behaviour of the system, and if it is determined that the quadratic approximation is suitable for the aforesaid purpose; (h) deriving from the said quadratic approximation parameters to serve as the basis for a measurement update for the modelling of the behaviour of the system; (i) subtracting the said quadratic approximation from the sum of the log-likelihood function and the residual function which it approximates so as to form a new residual function; and (j) otherwise using said sum of the log-likelihood function and the residual function used for determining said quadratic approximation as the new said residual function.

Claim 17

A method of measuring the state of a dynamic system comprising the steps of: (a) producing a sequence of measurements or sets of measurements which are dependent upon the current state of the said system in the presence of noise and other sources of error being sources which are representable by a statistical model; (b) determining a likelihood function associated with each of the said measurements or sets of measurements in accordance with the said statistical model; (c) obtaining the product of the said likelihood function and a residual function which is initially set to be identically unity; (d) transforming the said residual function to take into account dynamic change in the state of the system during the interval change in the state of the system during the interval between successive measurements or sets of measurements in the said measurement sequence; (e) utilising a recursive estimator to model the dynamic behaviour of the system; (f) determining a Gaussian approximation to the said product of the likelihood function and the residual function, the fit being weighted in dependence upon the present uncertainty in the state of the dynamic system as evidenced by the modelled behaviour of the system; (g) determining whether the said Gaussian approximation is suitable to serve as an input for said modelling of the system, and if it is determined that the Gaussian approximation is suitable for the aforesaid purpose; (h) deriving from the said Gaussian approximation parameters to serve as the basis for the modelling of the system; (i) dividing the said Gaussian approximation into the said product of the likelihood function and the residual function which it approximates so as to form a new residual function; and (j) otherwise using said product of the likelihood function and the residual function used for determining said Gaussian approximation as said new residual function.

Claim 18

A method of measuring the state of a dynamic system comprising the steps of: (a) producing a sequence of measurements or sets of measurements which are dependent upon the current state of the said system in the presence of noise and other sources of error being sources which are representable by a statistical model; (b) determining a likelihood function associated with each of the said measurements or sets of measurements in accordance with the said statistical model; (c) obtaining the product of the said likelihood function and a residual function which is initially set to be identically unity; (d) transforming the said residual function to take into account dynamic change in the state of the system during the interval change in the state of the system during the interval between successive measurements or sets of measurements in the said measurement sequence; (e) utilising a recursive estimator to model the dynamic behaviour of the system; (f) determining a Gaussian approximation to the said product of the likelihood function and the residual function, the fit being weighted in dependence upon the present uncertainty in the state of the dynamic system as evidenced by the modelled behaviour of the system; (g) determining whether the said Gaussian approximation is suitable to serve as an input for said modelling of the system, and if it is determined that the Gaussian approximation is suitable for the aforesaid purpose; (h) deriving from the said Gaussian approximation parameters to serve as the basis for the modelling of the system; (i) dividing the said Gaussian approximation into the said product of the likelihood function and the residual function which it approximates so as to form a new residual function; and (j) otherwise using said product of the likelihood function and the residual function used for determining said Gaussian approximation as said new residual function.

Claim 19

A method of using a recursive estimator for navigation integration in a navigation system wherein the parameters of a quadratic function approximating to a log-likelihood function are used as inputs to the estimator.

Claim 20

A method of using a recursive estimator for navigation integration in a navigation system wherein the parameters of a quadratic function approximating to a log-likelihood function are used as inputs to the estimator.

Claim 21

A method of using a recursive estimator for navigation integration in a navigation system wherein the parameters of a Gaussian function approximating to a likelihood function are used as inputs to the recursive estimator.

Claim 22

A method of using a recursive estimator for navigation integration in a navigation system wherein the parameters of a Gaussian function approximating to a likelihood function are used as inputs to the recursive estimator.