Position Estimation Under Multipath Transmission

• US 20200132861A1
• Filed: 10/31/2018
• Published: 04/30/2020
• Est. Priority Date: 10/31/2018
• Status: Active Grant

• Current Assignee: Mitsubishi Electric Research Laboratories, Inc. (Mitsubishi Electric Corporation)
• Original Assignee: Mitsubishi Electric Research Laboratories, Inc. (Mitsubishi Electric Corporation)
• Inventors: Kim, Kyeong Jin, Orlik, Philip, Hori, Chiori
• Granted Patent: US 11,079,495 B2
• Time in Patent Office: Days
• CPC Class Codes: G01S 19/22 Multipath-related issues
• CPC Class Codes: G01S 19/428 using multipath or indirect path propagation signals in position determination
• CPC Class Codes: G01S 19/43 using carrier phase measurements, e.g. kinematic positioning; using long or short baseline interferometry
• CPC Class Codes: G01S 19/50 whereby the position solution is constrained to lie upon a particular curve or surface, e.g. for locomotives on railway tracks
• CPC Class Codes: G01S 5/0218 Multipath in signal reception
• CPC Class Codes: G01S 5/02521 using a radio-map
• CPC Class Codes: G06N 3/044 Recurrent networks, e.g. Hopfield networks
• CPC Class Codes: G06N 3/045 Combinations of networks
• CPC Class Codes: G06N 3/048 Activation functions
• CPC Class Codes: G06N 3/049 Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
• CPC Class Codes: G06N 3/08 Learning methods

Abstract

A positioning system for tracking a position of a vehicle includes a receiver configured to receive phase measurements of satellite signals received at multiple instances of time from multiple satellites, and a memory configured to store a recurrent neural network trained to determine a position of the vehicle from a set of phase measurements in a presence of noise caused by a multipath transmission of at least some of the satellite signals at some instances of time. A processor of the positioning system is configured to track the position of the vehicle over different instances of time by processing the set of phase measurements received at each instance of time with the recurrent neural network to produce the position of the vehicle at each instance of time.

Claims

1. A positioning system for tracking a position of a vehicle, comprising:

a receiver configured to receive phase measurements of satellite signals received at multiple instances of time from multiple satellites;

a memory configured to store a recurrent neural network trained to determine a position of the vehicle from a set of phase measurements in a presence of noise caused by a multipath transmission of at least some of the satellite signals at some instances of time; and

a processor configured to track the position of the vehicle over different instances of time by processing the set of phase measurements received at each instance of time with the recurrent neural network to produce the position of the vehicle at each instance of time.

2. The positioning system of claim 1, wherein the recurrent neural network uses an attention-based multimodal fusion to output the position in response to receiving the set of phase measurements.

3. The positioning system of claim 1, wherein the recurrent neural network applies different weights to at least some different phase measurements to produce a set of weighted phase measurements and determines the position of the vehicle from the set of weighted phase measurements.

4. The positioning system of claim 1, wherein the recurrent neural network is trained to

determine individual weights of phase measurements of each satellite based on current and previous temporal features of the phase measurements of the satellite;

determine joint weights of combined individually weighted phase measurements of the multiple satellites based on previous joint weights; and

estimate the position of the vehicle from the combined individually weighted phase measurements that are weighted with the joint weights.

5. The positioning system of claim 4, wherein the recurrent neural network is further trained to extract current temporal features of the phase measurements of the satellite from a sequence of phase measurements collected over a current window of time;

retrieve previous temporal features of the phase measurements of the satellite determined for a sequence of phase measurements collected over a previous window of time and individual weights determined for the previous temporal features; and

determine the individual weights for the current temporal features that characterize temporal dependencies in the input phase measurements by adjusting the individual weights or temporal attention weights determined for the previous temporal features based on comparison of the current temporal features with the previous temporal features.

6. The positioning system of claim 4, wherein the recurrent neural network is further trained to

combine the individually weighted phase measurements received from the multiple satellites over a current window of time to produce the combined individually weighted phase measurements for the current window of time;

retrieve previously combined individually weighted phase measurements determined for a previous window of time;

retrieve previous joint weights determined for the previously combined individually weighted phase measurements; and

determine the joint weights that characterize temporal dependencies in the whole input phase measurements by adjusting previous joint weights based on comparison of the combined individually weighted phase measurements and the previously combined individually weighted phase measurements.

7. The positioning system of claim 1, wherein the recurrent neural network is trained for a specific path, such that the processor tracks the position of the vehicle traveling along at least a portion of the specific path.

8. The positioning system of claim 7, wherein the vehicle is a train traveling along the specific path.

9. The positioning system of claim 1, wherein the memory stores a set of recurrent neural networks, each recurrent neural network is trained for a specific path, wherein the processor selects the recurrent neural network from the set of recurrent neural networks based on the tracking.

10. The positioning system of claim 1, wherein the memory stores a set of recurrent neural networks, each recurrent neural network is trained for a specific path, wherein the processor uses different recurrent neural networks from the set of recurrent neural networks during the tracking.

11. The positioning system of claim 1, wherein the positioning system is a distributed system implemented over a vehicular network including the vehicle in communication with an access point of the vehicular network, wherein the recurrent neural network is stored in the memory of the access point and is trained to determined positions in an area of the vehicular network covered by the access point.

12. The positioning system of claim 11, wherein the processor uses different recurrent neural networks stored at different access points of the vehicular during the tracking.

13. The positioning system of claim 11, further comprising:

a transmitter to transmit the phase measurements to the access point.

14. The positioning system of claim 11, wherein the processor is running by the access point, further comprising:

a transceiver to transmit the phase measurements to the access point and to receive estimates of the position of the vehicle from the access point.

15. The positioning system of claim 1, further comprising:

a navigator configured to track the position of the vehicle to produce a log file including phase measurements and corresponding position estimates;

a trainer configured to train the recurrent neural network using data extracted from the log file; and

a switcher configured to switch the tracking from the navigator to the recurrent neural network when the recurrent neural network is trained.

16. The positioning system of claim 1, wherein the processor processes the set of phase measurements received at each instance of time with the recurrent neural network without testing the phase measurements for the multipath transmission.

17. A positioning method for tracking a position of a vehicle, wherein the method uses a processor coupled with stored instructions implementing the method, wherein the instructions, when executed by the processor carry out steps of the method, comprising:

receiving phase measurements of satellite signals received at multiple instances of time from multiple satellites;

accessing a recurrent neural network trained to determine a position of the vehicle from a set of phase measurements in a presence of noise caused by a multipath transmission of at least some of the satellite signals at some instances of time; and

tracking the position of the vehicle over different instances of time by processing the set of phase measurements received at each instance of time with the recurrent neural network to produce the position of the vehicle at each instance of time.

18. The positioning method of claim 17, wherein the recurrent neural network uses an attention-based multimodal fusion to output the position in response to receiving the set of phase measurements.

19. The positioning method of claim 17, wherein the recurrent neural network applies different weights to at least some different phase measurements to produce a set of weighted phase measurements and determines the position of the vehicle from the set of weighted phase measurements.

20. A non-transitory computer readable storage medium embodied thereon a program executable by a processor for performing a method, the method comprising:

receiving phase measurements of satellite signals received at multiple instances of time from multiple satellites;

accessing a recurrent neural network trained to determine a position of the vehicle from a set of phase measurements in a presence of noise caused by a multipath transmission of at least some of the satellite signals at some instances of time; and

tracking the position of the vehicle over different instances of time by processing the set of phase measurements received at each instance of time with the recurrent neural network to produce the position of the vehicle at each instance of time.