Local positioning system which operates based on reflected wireless signals

• US 20050122252A1
• Filed: 08/20/2004
• Published: 06/09/2005
• Est. Priority Date: 12/04/2003
• Status: Active Grant

• Current Assignee: Sony Corporation (Sony Group Corp.)
• Original Assignee: Sony Corporation (Sony Group Corp.)
• Inventors: Xu, Jin, Ma, Yugang, Sun, Xiaobing, Lu, Wei, Okada, Kanzo
• Granted Patent: US 7,400,291 B2
• Time in Patent Office: Days
• US Class Current: 342/118
• CPC Class Codes: G01S 13/46 Indirect determination of position data
• CPC Class Codes: G01S 2013/466 by Trilateration, i.e. two antennas or two sensors determine separately the distance to a target, whereby with the knowledge of the baseline length, i.e. the distance between the antennas or sensors, the position data of the target is determined
• CPC Class Codes: G01S 5/0273 using multipath or indirect path propagation signals in position determination

Abstract

A local positioning system is proposed for wirelessly locating an object using existing features within a static environment, such as walls, as the references for determining the position of the system. An antenna 16 attached to the object transmits RF signals which are reflected by the surroundings. During a training mode, the reflected signals are used to train a neural network 22, 43 to map the position of the object to the characteristics of the reflected signals. During a working mode, the trained neural network is to identify the position of the object based on reflected signals in working mode. Optionally, the reflected signals may be subject to a clustering process before input to the neural network.

Claims

1. A system comprising:

a transmitter for transmitting wireless signals in all directions within a range of directions, the range of directions including at least all directions in a plane;

a receiver for receiving echoes of the wireless signals from any direction in the range;

a processor for processing the received echoes to derive echo data signals indicative of the distance of the system to a plurality of reflective surfaces; and

an adaptive system for;

during a training stage, learning at least one association between the echo data signals and the position of the system based on training data composed of records of echo data signals and associated system position data, and

during a working stage, using the echo data signals and the learnt association to determine the position of the system.

2. A system according to claim 1 in which the processor includes a distance vector determination unit for providing the echo data signals comprising numbers which indicate the distances from the system to each of a number of respective reflective surfaces.

3. A system according to claim 1 in which the processor in includes a distance vector determination unit for providing the echo data signals comprising numbers which indicate the difference of the squares of the distances of the system to a respective pair of the reflective surfaces.

4. A system according to claim 1, in which the processor is arranged to modify the relationship between the echoes and the echo data signals to generate echo data signals comprising a plurality of numbers having linear relationships to position values of the system measured relative to Cartesian axes.

5. A system according to claim 1 which includes a clustering and classifying unit, the clustering and classifying unit being arranged:

during the training stage, to partition the echo data signals of the training data into clusters, the adaptive system learning an association for each cluster between the echo data signals of that cluster and the associated positions of the system; and

during the working stage, to determine which cluster the echo data signals belong to, whereby the adaptive system determines the position of the system using the learnt association of the corresponding cluster.

6. A system according to claim 5 in which the clustering and classifying unit includes a clustering unit adapted during the training stage to derive a cluster index indicative of a cluster associated with the echo data signals, and a classifying unit arranged during the training stage to learn associations between the echo data signals and the cluster indices and during the working phase to output a cluster index based on the echo data signals.

7. A system according to claim 6 in which the classifying unit is arranged to employ a learning vector quantization algorithm.

8. A system according to claim 7 in which the clustering unit during the training stage is arranged to take as input a vector comprising the echo data signals and the associated positions.

9. A system according to claim 7 in which the clustering unit during the training stage is arranged to take as input a vector comprising a linear transformation derived from the echo data signals and associated positions.

10. A system according to claim 1 in which the adaptive system includes at least one neural network.

11. A system according to claim 5 in which the adaptive system includes a neural network for each cluster, the neural network being trained using training data for the corresponding cluster to learn the association for echo data signals of the corresponding cluster.

12. A system according to claim 11 in which each neural network is a two-layer network.

13. A system according to claim 12 in which each neural network performs a linear function of its inputs.

14. A system according to claim 5 in which the adaptive system includes for each cluster a respective transformation unit adapted during the training stage to derive a respective linear transformation of its inputs.

15. A system according to claim 14 in which the transformation units derive the linear transformation by a least squares algorithm based on (i) vectors of distances as an input and (ii) positions associated with the echo data signals as outputs.

16. A system according to claim 1 in which the signals are RF wireless signals.

17. A system according to claim 16 in which the transmitter is an omni-directional antenna.

18. A method of determining the position of an object, the method comprising:

transmitting wireless signals from the object in all directions within a range of directions which includes at least all directions in a plane;

receiving at the object echoes of the wireless signals from any of the range of directions;

processing the received echoes to derive echo data signals indicative of the distance of the object to a plurality of reflective surfaces;

during a training stage, learning at least one association between the echo data signals and the position of the system based on echo data signals and associated object position data,

during a working stage, using the echo data signals and the learnt association to determine the position of the object.

19. A method according to claim 18 in which the echo data signals comprise numbers having linear relationships to position values of the object measured relative to Cartesian axes.

20. A method according to claim 18 in which the echo data comprise signals indicate the distances from the object to each of a number of respective reflective surfaces.

21. A method according to claim 19 in which the echo data comprise signals that indicate the distances from the object to each of a number of respective reflective surfaces and in which the respective reflective surfaces are substantially planar.

22. A method according to claim 18 in which the echo data signals comprise numbers which indicate the difference of the squares of the distances of the system to a respective pair of the reflective surfaces.

23. A method according to claim 19 in which the echo data signals comprise numbers which indicate the difference of the squares of the distances of the system to a respective pair of the reflective surfaces, and in which the respective reflective surfaces are point-like reflectors.

24. A method according to claim 19 including a step of determining a transform from the echoes to the echo data signals according to the nature of the reflective surfaces.

25. A method according to claim 18 further including:

during the training stage, partitioning echo data signals of the training data into clusters, the adaptive system learning an association for each cluster between the echo data signals of that cluster and the associated positions of the object; and

during the working stage, determining which cluster the echo data signals belong to, whereby the adaptive system determines the position of the object using the learnt association of the corresponding cluster.

26. A method according to claim 25 in which:

during the training stage a clustering unit derives a cluster index indicative of a cluster associated with the echo data signals, and a classifying unit learns associations between the echo data signals and the cluster indices, and

during the working phase to output a cluster index based on the echo data signals.

27. A method according to claim 26 in which the classifying unit employs a learning vector quantization algorithm.

28. A method according to claim 27 in which the clustering unit takes as input a vector comprising the echo data signals and the associated positions.

29. A method according to claim 27 in which the clustering unit takes as input a vector comprising a linear transformation derived from the echo data signals and associated positions.

30. A method according to claim 18 in which the learning is at least partly performed by at least one neural network.

31. A method according to claim 25 which employs a neural network for each cluster, the neural network being trained using training data for the corresponding cluster to learn the association for echo data signals of the corresponding cluster.

32. A method according to claim 31 in which each neural network is a two layer network.

33. A method according to claim 32 in which each neural network performs a linear function of its inputs.

34. A method according to claim 25 in which the adaptive system includes for each cluster a respective transformation unit which during the training stage derives a respective linear transformation of its inputs.

35. A method according to claim 34 in which the transformation units derive the linear transformation by a least squares algorithm based on (i) vectors of distances as an input and (ii) positions associated with the echo data signals as outputs.

36. A method according to claim 18 in which the range of directions is all directions within a three-dimensional space.