Method and system for predicting a location of an object in a multi-dimensional space

1. A method for predicting a location of an object in a multi-dimensional space having a plurality of (n+1) dimensions, the method comprising:

reading, by a computer system, an input probability density function defined by a set of parameters (μ,P,α,β,Γ,κ) and representing an uncertainty of the location of the object in a cylindrical multi-dimensional space ( n × ) of the multi-dimensional space and a diffeomorphism (Φ) of the cylindrical multi-dimensional space (Φ: n × → n × );

generating, by the computer system, a transformed probability density function defined by a set of parameters ({tilde over (μ)}, {tilde over (P)}, {tilde over (α)}, {tilde over (β)}, {tilde over (Γ)}, {tilde over (κ)}) and representing an uncertainty of the location of the object in a cylindrical multi-dimensional space, the transformed probability density function generated from the input probability density function under the diffeomorphism, wherein the input probability density function and the transformed probability density function are both represented by Gauss von Mises distributions defined on the cylindrical multi-dimensional space; and

providing, by the computer system, the set of parameters representing the transformed probability density function as a representation of the predicted location of the object in the multi-dimensional space.

2. The method of claim 1 , wherein generating the transformed probability density function comprises:

computing the parameters {tilde over (μ)} and {tilde over (P)} from a sequence of function evaluations Φ(x σ i ,θ σ i ), for i=1, . . . , N, where (x σ i ,θ σ i ), for i=1, . . . , N, are a chosen sequence of quadrature nodes on n × and wherein the quadrature nodes (x σ i ,θ σ i ), for i=1, . . . , N, are generated from a Gauss von Mises quadrature rule of a chosen order of accuracy in conjunction with the input parameter set (μ,P,α,β,Γ,κ);

selecting {tilde over (κ)}=κ;

approximating the parameters {tilde over (α)}, {tilde over (β)}, and {tilde over (Γ)} by {circumflex over (α)}, {circumflex over (β)}, and {circumflex over (Γ)}, respectively, using expressions depending on the partial derivatives of the diffeomorphism Φ;

selecting the parameters {tilde over (α)}, {tilde over (β)}, and {tilde over (Γ)} according to

(

α

~

,

β

~

,

Γ

~

)

=

arg

⁢

⁢

min

α

^

,

β

^

,

Γ

^

⁢

∑

i

=

1

N

⁢

⁢

[

M

⁡

(

x

σ

i

,

θ

σ

i

;

μ

,

P

,

α

,

β

,

Γ

,

κ

)

-

M

⁡

(

Φ

⁡

(

x

σ

i

,

θ

σ

i

)

;

μ

~

,

P

~

,

α

^

,

β

^

,

Γ

^

,

κ

~

)

]

2

,

where M is a Mahalanobis von Mises statistic.

3. The method of claim 1 , further comprising validating, by the computer system, the accuracy of the transformed Gauss von Mises distribution using a differential entropy metric.

4. The method of claim 1 , further comprising validating, by the computer system, the accuracy of the transformed Gauss von Mises distribution using a metric based on a Mahalanobis von Mises statistic.

5. The method of claim 1 , wherein the representation of the input and transformed probability density functions are mixtures of Gauss von Mises distributions defined on the multi-dimensional space.

6. The method of claim 1 , wherein the diffeomorphism is a solution flow induced from a system of ordinary differential equations describing two-body dynamics of orbital mechanics for the object.

7. The method of claim 1 , wherein the input Gauss von Mises distribution is generated from a plurality of radar, electro-optical, or infrared sensor observations.

8. The method of claim 1 , wherein the input Gauss von Mises distribution is generated from a Gaussian distribution defined on the multi-dimensional space.

9. The method of claim 1 , further comprising testing, by the computer system, if a point of the cylindrical multi-dimensional space is a statistically significant realization of a Gauss von Mises distribution.

10. The method of claim 1 , further comprising computing, by the computer system, the marginal distribution of a random vector from a Gauss von Mises distribution defined in the cylindrical multi-dimensional space.

11. A system comprising:

a processor; and

a memory coupled with and readable by the processor and having stored therein a sequence of instructions which, when executed by the processor, causes the processor to predict a location of an object in a multi-dimensional space having a plurality of (n+1) dimensions by:

receiving an input probability density function defined by a set of parameters (μ,P,α,β,Γ,κ) and representing an uncertainty of the location of the object in a cylindrical multi-dimensional space ( n × ) of the multi-dimensional space and a diffeomorphism (Φ) of the cylindrical multi-dimensional space (Φ: n × → n × ),

generating a transformed probability density function defined by a set of parameters ({tilde over (μ)}, {tilde over (P)}, {tilde over (α)}, {tilde over (β)}, {tilde over (Γ)}, {tilde over (κ)}) and representing an uncertainty of the location of the object in a cylindrical multi-dimensional space, the transformed probability density function generated from the input probability density function under the diffeomorphism, wherein the input probability density function and the transformed probability density function are both represented by Gauss von Mises distributions defined on the cylindrical multi-dimensional space, and

providing the set of parameters representing the transformed probability density function as a representation of the predicted location of the object in the multi-dimensional space.

12. The system of claim 11 , wherein generating the transformed probability density function comprises:

computing the parameters {tilde over (μ)} and {tilde over (P)} from a sequence of function evaluations Φ(x σ i ,θ σ i ), for i=1, . . . , N, where (x σ i ,θ σ i ), for i=1, . . . , N, are a chosen sequence of quadrature nodes on n × and wherein the quadrature nodes (x σ i ,θ σ i ), for i=1, . . . , N, are generated from a Gauss von Mises quadrature rule of a chosen order of accuracy in conjunction with the input parameter set (μ,P,α,β,Γ,κ);

selecting {tilde over (κ)}=κ;

approximating the parameters {tilde over (α)}, {tilde over (β)}, and {tilde over (Γ)} by {circumflex over (α)}, {circumflex over (β)}, and {circumflex over (Γ)}, respectively, using expressions depending on the partial derivatives of the diffeomorphism Φ;

selecting the parameters {tilde over (α)}, {tilde over (β)}, and {tilde over (Γ)} according to

(

α

~

,

β

~

,

Γ

~

)

=

arg

⁢

⁢

min

α

^

,

β

^

,

Γ

^

⁢

∑

i

=

1

N

⁢

⁢

[

M

⁡

(

x

σ

i

,

θ

σ

i

;

μ

,

P

,

α

,

β

,

Γ

,

κ

)

-

M

⁡

(

Φ

⁡

(

x

σ

i

,

θ

σ

i

)

;

μ

~

,

P

~

,

α

^

,

β

^

,

Γ

^

,

κ

~

)

]

2

,

where M is a Mahalanobis von Mises statistic.

13. The system of claim 11 , further comprising validating the accuracy of the transformed Gauss von Mises distribution using a differential entropy metric.

14. The system of claim 11 , further comprising validating the accuracy of the transformed Gauss von Mises distribution using a metric based on a Mahalanobis von Mises statistic.

15. The system of claim 11 , wherein the representation of the input and transformed probability density functions are mixtures of Gauss von Mises distributions defined on the multi-dimensional space.

16. The system of claim 11 , wherein the diffeomorphism is a solution flow induced from a system of ordinary differential equations describing two-body dynamics of orbital mechanics for the object.

17. The system of claim 11 , wherein the input Gauss von Mises distribution is generated from a plurality of radar, electro-optical, or infrared sensor observations.

18. The system of claim 11 , wherein the input Gauss von Mises distribution is generated from a Gaussian distribution defined on the multi-dimensional space.

19. The system of claim 11 , further comprising testing if a point of the cylindrical multi-dimensional space is a statistically significant realization of a Gauss von Mises distribution.

20. The system of claim 11 , further comprising computing the marginal distribution of a random vector from a Gauss von Mises distribution defined in the cylindrical multi-dimensional space.

21. A computer-readable medium having stored thereon a sequence of instructions which, when executed by a processor, causes the processor to predict a location of an object in a multi-dimensional space having a plurality of (n+1) dimensions by:

receiving an input probability density function defined by a set of parameters (μ,P,α,β,Γ,κ) and representing an uncertainty of the location of the object in a cylindrical multi-dimensional space ( n × ) of the multi-dimensional space and a diffeomorphism (Φ) of the cylindrical multi-dimensional space (Φ: n × → n × );

generating a transformed probability density function defined by a set of parameters ({tilde over (μ)}, {tilde over (P)}, {tilde over (α)}, {tilde over (β)}, {tilde over (Γ)}, {tilde over (κ)}) and representing an uncertainty of the location of the object in a cylindrical multi-dimensional space, the transformed probability density function generated from the input probability density function under the diffeomorphism, wherein the input probability density function and the transformed probability density function are both represented by Gauss von Mises distributions defined on the cylindrical multi-dimensional space; and

providing the set of parameters representing the transformed probability density function as a representation of the predicted location of the object in the multi-dimensional space.

22. The computer-readable medium of claim 21 , wherein generating the transformed probability density function comprises:

computing the parameters {tilde over (μ)} and {tilde over (P)} from a sequence of function evaluations Φ(x σ i ,θ σ i ), for i=1, . . . , N, where (x σ i ,θ σ i ), for i=1, . . . , N, are a chosen sequence of quadrature nodes on n × and wherein the quadrature nodes (x σ i ,θ σ i ), for i=1, . . . , N, are generated from a Gauss von Mises quadrature rule of a chosen order of accuracy in conjunction with the input parameter set (μ,P,α,β,Γ,κ);

selecting {tilde over (κ)}=κ;

approximating the parameters {tilde over (α)}, {tilde over (β)}, and {tilde over (Γ)} by {circumflex over (α)}, {circumflex over (β)}, and {circumflex over (Γ)}, respectively, using expressions depending on the partial derivatives of the diffeomorphism Φ;

selecting the parameters {tilde over (α)}, {tilde over (β)}, and {tilde over (Γ)} according to

(

α

~

,

β

~

,

Γ

~

)

=

arg

⁢

⁢

min

α

^

,

β

^

,

Γ

^

⁢

∑

i

=

1

N

⁢

⁢

[

M

⁡

(

x

σ

i

,

θ

σ

i

;

μ

,

P

,

α

,

β

,

Γ

,

κ

)

-

M

⁡

(

Φ

⁡

(

x

σ

i

,

θ

σ

i

)

;

μ

~

,

P

~

,

α

^

,

β

^

,

Γ

^

,

κ

~

)

]

2

,

where M is a Mahalanobis von Mises statistic.

23. The computer-readable medium of claim 21 , further comprising validating the accuracy of the transformed Gauss von Mises distribution using a differential entropy metric.

24. The computer-readable medium of claim 21 , further comprising validating the accuracy of the transformed Gauss von Mises distribution using a metric based on a Mahalanobis von Mises statistic.

25. The computer-readable medium of claim 21 , wherein the representation of the input and transformed probability density functions are mixtures of Gauss von Mises distributions defined on the multi-dimensional space.

26. The computer-readable medium of claim 21 , wherein the diffeomorphism is a solution flow induced from a system of ordinary differential equations describing two-body dynamics of orbital mechanics for the object.

27. The computer-readable medium of claim 21 , wherein the input Gauss von Mises distribution is generated from a plurality of radar, electro-optical, or infrared sensor observations.

28. The computer-readable medium of claim 21 , wherein the input Gauss von Mises distribution is generated from a Gaussian distribution defined on the multi-dimensional space.

29. The computer-readable medium of claim 21 , further comprising testing if a point of the cylindrical multi-dimensional space is a statistically significant realization of a Gauss von Mises distribution.

30. The computer-readable medium of claim 21 , further comprising computing the marginal distribution of a random vector from a Gauss von Mises distribution defined in the cylindrical multi-dimensional space.