Neural network-based radar system having independent multibeam antenna

• US 10,365,350 B2
• Filed: 01/28/2016
• Issued: 07/30/2019
• Est. Priority Date: 01/29/2015
• Status: Active Grant

• Current Assignee: National University Corporation Shizuoka University, Nidec Corporation
• Original Assignee: Nidec Corporation
• Inventors: Kamo, Hiroyuki, Kuwahara, Yoshihiko
• Primary Examiner(s): McGue, Frank J
• Application Number:
  US15/008,624
• Publication Number:
  US 20160223651A1
• Time in Patent Office: 1,279 Days
• Field of Search: 342 70
• CPC Class Codes: G01S 13/426 Scanning radar, e.g. 3D radar G01S13/66 takes precedence
• CPC Class Codes: G01S 13/726 Multiple target tracking
• CPC Class Codes: G01S 13/931 of land vehicles
• CPC Class Codes: G01S 2013/93185 Controlling the brakes
• CPC Class Codes: G01S 2013/9319 Controlling the accelerator
• CPC Class Codes: G01S 2013/9321 Velocity regulation, e.g. cruise control
• CPC Class Codes: G01S 7/417 involving the use of neural networks

Abstract

A radar system includes an independent multibeam antenna which outputs at least one reception signal in response to at least one arriving wave, and a signal processing circuit in which a learned neural network has been established. The signal processing circuit receives the at least one reception signal, inputs the at least one reception signal or a secondary signal generated from the at least one reception signal to the neural network, performs computation by using the at least one reception signal or secondary signal and learned data of the neural network, and outputs a signal indicating the number of at least one arriving wave from the neural network.

Claims

1. A radar system comprising:

an independent multibeam antenna which outputs at least one reception signal in response to at least one arriving wave; and

a signal processing circuit in which a learned neural network has been established;

wherein

the signal processing circuit includes;

a neural network circuit in which the learned neural network is implemented, and

an arriving wave estimation circuit to estimate an angle or angles indicating a direction or directions of arrival of the at least one arriving wave and output a signal indicating a result of estimation; and

the signal processing circuit;

receives the at least one reception signal from the independent multibeam antenna;

inputs the at least one reception signal or a secondary signal generated from the at least one reception signal, to the learned neural network;

performs, in the neural network circuit, computation by using the at least one reception signal or secondary signal and data of the learned neural network; and

outputs a signal indicating a number of the at least one arriving wave from the neural network, and

receives, in the arriving wave estimation circuit, the at least one reception signal or secondary signal, and the signal indicating a number of the at least one arriving wave from the neural network circuit.

2. The radar system of claim 1, wherein the independent multibeam antenna includes a lens or a reflector and a plurality of antenna elements.

3. The radar system of claim 1, wherein the independent multibeam antenna includes one or more antenna elements and an actuator to alter the position of the one or more antenna elements to achieve directivity control at least concerning beam reception, the actuator mechanically driving a part or all of the antenna elements so that the number of plural beams to be received in different directions is greater than the number of antenna elements.

4. The radar system of claim 1, wherein the signal processing circuit is configured to estimate a direction or directions of arrival of the at least one arriving wave with a maximum likelihood estimation technique by using the signal indicating the number of the at least one arriving wave, and output an estimated value.

18. The radar system of claim 1, wherein the signal processing circuit detects at least one target based on the plurality of reception signals, and determines a distance to the at least one target and a relative velocity of the at least one target relative to the array antenna.

21. The radar system of claim 1, wherein the signal processing circuit includes:

only one signal processing circuit to perform the computation; and

a memory device storing the data of the learned neural network.

22. The radar system of claim 1, wherein the data of the learned neural network includes values of weights to be applied to inputs to nodes of the learned neural network.

5. A radar system comprising:

an independent multibeam antenna which includes a plurality of antenna elements and outputs a plurality of reception signals in response to at least one arriving wave; and

a signal processing circuit in which a learned neural network has been established;

wherein

the signal processing circuit;

receives the plurality of reception signals;

inputs a unit vector generated from the plurality of reception signals, to the neural network;

performs computation by using the unit vector, and data of the learned neural network; and

outputs a signal indicating a number of the at least one arriving wave from the neural network;

the signal processing circuit is configured to estimate a direction or directions of arrival of the at least one arriving wave with a maximum likelihood estimation technique by using the signal indicating the number of the at least one arriving wave, and output an estimated value;

the signal processing circuit further includes

a neural network circuit in which the neural network is implemented, and

an arriving wave estimation circuit to estimate an angle or angles indicating the direction or directions of arrival of the at least one arriving wave and output a signal indicating a result of estimation; and

the arriving wave estimation circuit receives the unit vector and the signal indicating the number of the at least one arriving wave from the neural network circuit, and performs computation of the maximum likelihood estimation technique by using the number of the at least one arriving wave and complex amplitude data based on the unit vector, thereby detecting angles indicating the direction or directions of arrival of the at least one arriving wave.

6. The radar system of claim 5, wherein,

the neural network circuit and the arriving wave estimation circuit receive a beat signal generated based on the plurality of reception signals;

and

the arriving wave estimation circuit is configured to

generate a covariance matrix from the complex amplitude data,

apply eigenvalue decomposition to the covariance matrix, and

calculate an angle that possesses a largest likelihood from at least one eigenvalue and at least one eigenvector.

7. The radar system of claim 6, wherein,

the arriving wave estimation circuit is configured to

calculate distance and relative velocity of a wave source or wave sources of the at least one arriving wave by using a frequency component of the beat signal, and

establish at least one of the wave sources at least based on the distance and relative velocity, and the angle that possesses the largest likelihood.

8. The radar system of claim 5, wherein

the radar system is mounted in a driver'"'"'s vehicle;

and

the signal processing circuit outputs the signal indicating the number of the at least one arriving wave or a signal indicating a spatial distribution pattern of preceding vehicles, as a signal indicating a number of preceding vehicles ahead of the driver'"'"'s vehicle, where the preceding vehicles are one preceding vehicle or a plurality of the preceding vehicles that are abreast of each other.

9. The radar system of claim 8, wherein as a signal indicating a spatial distribution pattern of the one preceding vehicle or the plurality of preceding vehicles that are abreast of each other, the signal processing circuit outputs a signal indicating whether or not at least one of the preceding vehicles exists in a driver'"'"'s lane in which the driver'"'"'s vehicle is travelling and in a lane adjacent to the driver'"'"'s lane.

10. The radar system of claim 9, wherein the signal processing circuit outputs a signal indicating a spatial distribution pattern of the one preceding vehicle or the plurality of preceding vehicles that are abreast of each other in terms of numerical values.

11. The radar system of claim 9, wherein the signal processing circuit operates in an output mode of outputting a signal indicating a spatial distribution pattern of the one preceding vehicle or the plurality of preceding vehicles that are abreast of each other without estimating a direction or directions of arrival of the at least one arriving wave.

12. The radar system of claim 11, wherein

the signal processing circuit further operates in an estimation mode of estimating a direction or directions of arrival of the at least one arriving wave without using the signal indicating the number of the least one arriving wave;

and

the signal processing circuit acquires a spatial distribution pattern of the one preceding vehicle or the plurality of preceding vehicles in the output mode, and thereafter performs processing in the estimation mode by utilizing the spatial distribution pattern of the one preceding vehicle or the plurality of preceding vehicles.

13. The radar system of claim 9, wherein the learned neural network completes learning in a situation where other vehicles respectively exist in two adjoining lanes of a same direction as a direction of travel of the driver'"'"'s vehicle and in a situation where none of the other vehicles exist, by using the plurality of reception signals obtained in each situation and by using a training signal indicating a spatial distribution of vehicles in the lanes in each situation.

14. The radar system of claim 8, wherein the signal processing circuit estimates a direction or directions of arrival of the at least one arriving wave based on the plurality of reception signals.

15. The radar system of claim 14, wherein the signal processing circuit selectively operates in one of:

a first mode of estimating a direction or directions of arrival of the at least one arriving wave by using the signal indicating the number of the at least one arriving wave output from the learned neural network; and

a second mode of estimating a direction or directions of arrival of the at least one arriving wave without using the signal indicating the number of the at least one arriving wave.

16. The radar system of claim 15, wherein the second mode estimates a direction or directions of arrival of the at least one arriving wave by performing an algorithm for direction-of-arrival estimation.

17. The radar system of claim 8, wherein

the signal processing circuit uses a camera or radio wave to identify a lane in which the driver'"'"'s vehicle is traveling;

the learned neural network has been established in a lane-by-lane manner for predefined lanes for the driver'"'"'s vehicle to travel in; and

the signal processing circuit selects the learned neural network in accordance with the lane in which the driver'"'"'s vehicle is traveling, and outputs a signal indicating a spatial distribution pattern of the one preceding vehicle or the plurality of preceding vehicles.

19. The radar system of claim 5, wherein the signal processing circuit is a programmable logic device.

20. The radar system of claim 19, wherein the signal processing circuit includes:

a plurality of computation elements to perform the computation; and

a plurality of memory elements storing the data of the learned neural network.

23. A radar signal processing apparatus to be mounted in a driver'"'"'s vehicle, the radar signal processing apparatus comprising:

a first terminal to acquire, from an independent multibeam antenna which outputs at least one reception signal in response to at least one arriving wave, the at least one reception signal;

a signal processing circuit in which a learned neural network has been established; and

a second terminal to transmit a signal which is output from the signal processing circuit;

wherein

the signal processing circuit includes;

a neural network circuit in which the learned neural network is implemented, and

an arriving wave estimation circuit to estimate an angle or angles indicating a direction or directions of arrival of the at least one arriving wave and output a signal indicating a result of estimation; and

the signal processing circuit;

receives the at least one reception signal via the first terminal;

inputs the at least one reception signal or a secondary signal generated from the at least one reception signal to the learned neural network;

performs, in the neural network circuit, computation by using the at least one reception signal or secondary signal and data of the learned neural network;

outputs a signal indicating a number of the at least one arriving wave; and

receives, in the arriving wave estimation circuit, the at least one reception signal or secondary signal, and the signal indicating a number of the at least one arriving wave from the neural network circuit.

24. The radar signal processing apparatus of claim 23, wherein the signal processing circuit further outputs, as computed by using the at least one reception signal, a signal indicating a distance between a driver'"'"'s vehicle and at least one preceding vehicle in a driver'"'"'s lane and in a lane adjacent to the driver'"'"'s lane, and a signal indicating a relative velocity of the at least one preceding vehicle.

25. A vehicle travel controlling apparatus to be mounted in a driver'"'"'s vehicle, the vehicle travel controlling apparatus comprising:

an independent multibeam antenna which outputs at least one reception signal in response to at least one arriving wave;

the radar signal processing apparatus of claim 24; and

an electronic control apparatus to perform adaptive cruise control of the driver'"'"'s vehicle and control a velocity of the driver'"'"'s vehicle by using signals being output from the radar signal processing apparatus and respectively indicating a distance, a relative velocity, and a number of the at least one arriving wave.

26. A vehicle travel controlling method comprising:

acquiring at least one reception signal by using an independent multibeam antenna which outputs at least one reception signal in response to at least one arriving wave;

inputting the at least one reception signal or a secondary signal generated from the at least one reception signal to a signal processing circuit in which a learned neural network has been established;

performing computation in the signal processing circuit by using the at least one reception signal or secondary signal and data of the learned neural network;

outputting a signal indicating a number of the at least one arriving wave from the learned neural network;

by using the at least one reception signal, outputting signals respectively indicating distances between a driver'"'"'s vehicle and at least one preceding vehicle in a driver'"'"'s lane and in a lane adjacent to the driver'"'"'s lane as corresponding to the at least one arriving wave and a relative velocity of the at least one preceding vehicle; and

performing adaptive cruise control of the driver'"'"'s vehicle by using the signals respectively indicating a distance, a relative velocity, and a number of the at least one arriving wave to control a velocity of the driver'"'"'s vehicle.

27. A non-transitory computer-readable medium having stored thereon a computer program for a vehicle identification process to be executed by a signal processing circuit mounted in a vehicle,

the vehicle including an independent multibeam antenna which outputs at least one reception signal in response to at least one arriving wave, and

the signal processing circuit includes:

a neural network circuit in which the learned neural network is implemented, and

an arriving wave estimation circuit to estimate an angle or angles indicating a direction or directions of arrival of the at least one arriving wave and output a signal indicating a result of estimation; and

the computer program causing the signal processing circuit to;

receive the at least one reception signal from the independent multibeam antenna;

input the at least one reception signal or a secondary signal generated from the at least one reception signal to the learned neural network;

perform computation by using the at least one reception signal or secondary signal and data of the learned neural network; and

output a signal indicating a number of the at least one arriving wave from the learned neural network.