Autonomous navigation apparatus with neural network for a mobile vehicle

• US 5,548,512 A
• Filed: 10/04/1994
• Issued: 08/20/1996
• Est. Priority Date: 10/04/1994
• Status: Expired due to Fees

• Current Assignee: United States Incorporated
• Original Assignee: United States Of America As Represented By The Administrator Of The National Aeronautics And Space Administration
• Inventors: Quraishi, Naveed
• Primary Examiner(s): ZANELLI, MICHAEL J
• Application Number:
  US08/317,443
• Time in Patent Office: 686 Days
• Field of Search: 364/424.02, 364/460, 364/434, 364/559, 180/167, 180/168, 180/169, 318/587, 395/94, 395/905, 395/86, 395/23, 395/24, 395/93
• US Class Current: 701/23
• CPC Class Codes: G05D 1/0221 involving a learning process
• CPC Class Codes: G05D 1/0255 using acoustic signals, e.g. ultra-sonic singals sonar systems designed for anti-collision purposes G01S15/93
• CPC Class Codes: G05D 1/0272 comprising means for registering the travel distance, e.g. revolutions of wheels measuring distance traversed on the ground by vehicles, e.g. using odometers G01C22/00
• CPC Class Codes: G06N 3/045 Combinations of networks
• CPC Class Codes: Y10S 706/905 Vehicle or aerospace

Abstract

An autonomous navigation system for a mobile vehicle arranged to move within an environment includes a plurality of sensors arranged on the vehicle and at least one neural network including an input layer coupled to the sensors, a hidden layer coupled to the input layer, and an output layer coupled to the hidden layer. The neural network produces output signals representing respective positions of the vehicle, such as the X coordinate, the Y coordinate, and the angular orientation of the vehicle. A plurality of patch locations within the environment are used to train the neural networks to produce the correct outputs in response to the distances sensed.

Claims

1. Autonomous navigation apparatus for a mobile vehicle arranged to move within an environment that is delimited, at least in part, by a boundary, said apparatus comprising, in combination:

(a) a plurality of sensors arranged on said vehicle and adapted to determine the distance from said vehicle to said boundary, each sensor being operative to determine said distance in a prescribed direction, with respect to said vehicle, which is different from the direction of at least one other sensor;

(b) a neural network comprising an input layer coupled to said sensors, at least one hidden layer coupled to said input layer and an output layer coupled to one of said hidden layers for producing a first output representing an identification of the spacial position of said vehicle; and

at least one patch location within said environment for training said neural network to produce a correct first output in response to the distance from said vehicle to said boundary determined by said plurality of sensors and the actual position of said vehicle at the at least one patch location.

2. The apparatus defined in claim 1, wherein the output layer of said neural network further produces a second output representing an identification of the spacial orientation of said vehicle.

3. The apparatus defined in claim 1, wherein said first output represents at least two orthogonal coordinates defining said spacial position.

4. The apparatus defined in claim 3, wherein said first output represents three orthogonal coordinates defining said spacial position.

5. The apparatus defined in claim 3, wherein said first output represents X and Y coordinates of said vehicle.

6. The apparatus defined in claim 1, further comprising means, connected to said input layer, for determining the motion of said vehicle with respect to said boundary.

7. The apparatus defined in claim 1, further comprising:

said mobile vehicle being adapted to move in a forward direction within the environment and having a central vertical axis perpendicular to said forward direction and a central horizontal axis which is parallel to said forward direction and which passes through said vertical axis; and

each of said plurality of sensors being operative to determine said distance in a prescribed angular direction, substantially perpendicular to said vertical axis of said vehicle, which is different from the angular direction of at least one other of the plurality of sensors.

8. The apparatus as defined in claim 7, wherein said neural network produces said first output representing a first spacial coordinate and a second output representing a second spacial coordinate.

9. The apparatus as defined in claim 8, wherein said first and second coordinates are Cartesian coordinates of the position of said vehicle.

10. The apparatus as defined in claim 1, further comprising:

at least one encoder coupled to said input layer of said neural network.

11. The apparatus as defined in claim 1, wherein said neural network comprises:

a first neural network including an input layer coupled to said sensors, a hidden layer coupled to said input layer and an output layer coupled to said hidden layer for producing a first output representing a first identification of the position of said vehicle;

a second neural network including an input layer coupled to said sensors, a hidden layer coupled to said input layer and an output layer coupled to said hidden layer for producing a second output representing a second identification of the position of said vehicle; and

each of said layers in said first and second networks comprise a plurality of network elements and wherein at least two of said layers of said first and second neural networks overlap such that a plurality of said network elements in said two layers are included in both said first and second neural networks.

12. The apparatus defined in claim 11, further comprising:

a third neural network comprising an input layer coupled to said sensors, a hidden layer coupled to said input layer and an output layer coupled to said hidden layer for producing a third output representing a third identification of the position of said vehicle; and

said input layer of said third neural network is coupled to at least one of said first and second outputs.

13. The apparatus defined in claim 12, wherein said first and second identifications are Cartesian coordinates of the said position of said vehicle and said third identification is an angular coordinate.

14. The apparatus defined in claim 12, wherein each of said layers in said first, second and third networks comprise a plurality of network elements and wherein at least two of said layers of at least two of said first, second and third neural networks overlap such that a plurality of said network elements in said two layers are included in more than one of said first, second and third neural networks.

15. Method of calibrating autonomous navigation apparatus for a mobile vehicle arranged to move in a forward direction within an environment that is delimited, at least in part, by a boundary, said vehicle having a central vertical axis perpendicular to said forward direction and a central horizontal axis which is parallel to said forward direction and which passes through said vertical axis, wherein said apparatus comprises, in combination:

(a) a plurality of sensors arranged on said vehicle and adapted to determine the distance from said vehicle to said boundary, each sensor being operative to determine said distance in a prescribed angular direction, perpendicular to said vertical axis of said vehicle, which is different from the angular direction of at least one other sensor;

(b) a first neural network comprising an input layer coupled to said sensors, a hidden layer coupled to said input layer and an output layer coupled to said hidden layer for producing a first output representing a first identification of the position of said vehicle; and

(c) a second neural network comprising an input layer coupled to said sensors, a hidden layer coupled to said input layer and an output layer coupled to said hidden layer for producing a second output representing a second identification of the position of said vehicle;

said calibration method comprising the steps of;

(1) selecting a plurality of patch locations within said environment;

(2) sensing the distances of said vehicle from said boundary using said plurality of sensors while said vehicle is disposed at each of said patch locations;

(3) determining the actual position of said vehicle at each patch location; and

(4) training said neural networks to produce correct first and second outputs in response to the distances sensed in step (2).

16. The method defined in claim 15, further comprising the steps of:

(5) rotating said vehicle about said vertical axis at each of said patch locations; and

(6) carrying out steps (2) through (4) at each angular orientation.

17. A method of calibrating the autonomous navigation apparatus for a mobile vehicle arranged to move in a forward direction within an environment that is delimited, at least in part, by a boundary, said vehicle having a central vertical axis perpendicular to said forward direction and a central horizontal axis which is parallel to said forward direction and which passes through said vertical axis, the method comprising:

arranging a plurality of sensors on said vehicle to determine the distance from said vehicle to said boundary;

positioning a neural network on said vehicle, said neural network including an input layer coupled to said plurality of sensors, a hidden layer coupled to said input layer and an output layer coupled to said hidden layer;

selecting at least one patch location within said environment;

determining the actual position of said vehicle at said at least one patch location;

while said vehicle is at said at least one patch location, determining said distance, for each of said plurality of sensors, from said vehicle to said boundary in a prescribed angular direction substantially perpendicular to said central axis of said vehicle; and

training said neural network in response to said sensed distances.

18. The method defined in claim 17, further comprising:

producing a first output from said neural network representing a first identification of the position of the said vehicle within said environment; and

producing a second output from said vehicle representing a second identification of the position of said vehicle within said environment.

19. The method defined in claim 17, further comprising:

sensing travel distance of said vehicle moving within said environment; and

inputting a signal representative of said sensed travel distance to said neural network.

20. The method defined in claim 17, further comprising:

rotating said mobile vehicle about said vertical axis within said at least one patch location; and

determining said distance at different angular orientations of said vehicle within said at least one patch location.

21. The method defined in claim 17, further comprising:

providing a plurality of spaced apart patch locations within said environment; and

determining said distance from said vehicle to said boundary at each of said patch locations for training said neural network.

22. The method defined in claim 21, further comprising:

maneuvering said vehicle between said spaced apart patch locations; and

training said neural network at each of said plurality of spaced apart patch locations.