Feature detection apparatus and methods for training of robotic navigation

• US 9,630,318 B2
• Filed: 11/14/2014
• Issued: 04/25/2017
• Est. Priority Date: 10/02/2014
• Status: Expired due to Fees

• Current Assignee: Brain Corporation
• Original Assignee: Brain Corporation
• Inventors: Ibarz Gabardos, Borja, Smith, Andrew, O'Connor, Peter
• Primary Examiner(s): Tran, Khoi
• Assistant Examiner(s): RINK, RYAN J
• Application Number:
  US14/542,391
• Publication Number:
  US 20160096270A1
• Time in Patent Office: 893 Days
• CPC Class Codes: B25J 9/0081 with master teach-in means
• CPC Class Codes: B25J 9/1602 characterised by the control system, structure, architecture
• CPC Class Codes: B25J 9/1607 Calculation of inertia, jacobian matrixes and inverses
• CPC Class Codes: B25J 9/161 Hardware, e.g. neural networks, fuzzy logic, interfaces, processor
• CPC Class Codes: B25J 9/163 learning, adaptive, model based, rule based expert control
• CPC Class Codes: B25J 9/1666 Avoiding collision or forbidden zones
• CPC Class Codes: B25J 9/1697 Vision controlled systems
• CPC Class Codes: G05D 1/0088 characterized by the autonomous decision making process, e.g. artificial intelligence, predefined behaviours using knowledge based models G06N5/00
• CPC Class Codes: G05D 1/0246 using a video camera in combination with image processing means
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 3/00 Computing arrangements based on biological models
• CPC Class Codes: G06N 3/008 based on physical entities controlled by simulated intelligence so as to replicate intelligent life forms, e.g. based on robots replicating pets or humans in their appearance or behaviour
• CPC Class Codes: G06N 3/049 Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
• CPC Class Codes: Y10S 901/01 Mobile robot
• CPC Class Codes: Y10S 901/03 Teaching system
• CPC Class Codes: Y10S 901/09 Closed loop, sensor feedback controls arm movement
• CPC Class Codes: Y10S 901/47 Optical

Abstract

A robotic device may be operated by a learning controller comprising a feature learning configured to determine control signal based on sensory input. An input may be analyzed in order to determine occurrence of one or more features. Features in the input may be associated with the control signal during online supervised training. During training, learning process may be adapted based on training input and the predicted output. A combination of the predicted and the target output may be provided to a robotic device to execute a task. Feature determination may comprise online adaptation of input, sparse encoding transformations. Computations related to learning process adaptation and feature detection may be performed on board by the robotic device in real time thereby enabling autonomous navigation by trained robots.

Claims

1. A method of determining a motor control signal for a robot based on an analysis of a sensor input, the method comprising:

analyzing the sensor input based on a first transformation process applied based on a presence of one or more first features present in the sensor input with a first characteristic;

analyzing the sensor input based on a second transformation process applied based on a presence of one or more second features present in the sensor input with a second characteristic, the second characteristic being different from the first characteristic;

determining a feature history comprising a plurality of features determined based on at least the acts of analyzing the sensor input based on the first transformation process over a first time interval and analyzing the sensor input based on the second transformation process over a second time interval;

normalizing each of the plurality of features using a statistical parameter determined from the feature history;

providing a plurality of the normalized features to a classification process for detecting at least one relevant feature among the plurality of the normalized features, the at least one relevant feature being characterized by a relevancy parameter that is larger than a predetermined threshold;

selecting from a plurality of control signals a motor control signal associated with the at least one relevant feature; and

executing the selected motor control signal, thereby causing a target task to be performed by the robot;

wherein the act of selecting the motor control signal is effectuated based on a prior contemporaneous occurrence of the motor control signal and the at least one relevant feature.

2. The method of claim 1, wherein the relevancy parameter larger than a predetermined threshold is generated according to one or more of a correlation, a probability, and a distance measure between an occurrence of the plurality of features and an occurrence of the motor control signal.

3. The method of claim 2, wherein:

the act of analyzing the sensor input with the first characteristic comprises analyzing;

a feature color, a feature texture, a feature edge orientation, and a feature type; and

the act of analyzing the sensor input with the second characteristic comprises analyzing a feature motion.

4. The method of claim 1, wherein:

the first transformation process comprises a sparsification operation comprising selecting one or more first features from a plurality of first outputs produced by the first transformation process based on the first characteristic; and

the sparsification operation further comprises reducing a number of data elements by a factor of at least four.

5. The method of claim 4, wherein the first characteristic comprises a statistical parameter further determined based on the plurality of first outputs of the first transformation process.

6. An apparatus configured to determine a motor control signal for a robot based on an analysis of an input signal, the apparatus comprising:

a processor; and

a non-transitory computer-readable medium configured to store at least one computer program thereon, the computer program comprising a plurality of instructions configured to, when executed by the processor, cause the apparatus to;

analyze the input signal based on a first transformation process applied based on a presence of at least one first feature from the input signal with a first characteristic;

analyze the input signal based on a second transformation process applied based on a presence of at least one second feature from the input signal with a second characteristic, the second characteristic being different from the first characteristic;

determine a plurality of features based on the analysis of the input signal based on the first transformation process over a first time interval and the analysis of the input signal based on the second transformation process over a second time interval;

normalize each of the plurality of features using a statistical parameter derived from the plurality of features;

provide a plurality of normalized features to a classification process configured to detect at least one relevant feature among the plurality of normalized features, the at least one relevant feature comprising a relevance parameter that exceeds a predetermined threshold;

select from a plurality of control signals a motor control signal associated with the at least one relevant feature with the relevance parameter that exceeds the predetermined threshold; and

apply the selected motor control signal to the robot, thereby causing the robot to execute a target task;

wherein the selection of the motor control signal is based on a prior simultaneous occurrence of the motor control signal and the at least one relevant feature.

7. The apparatus of claim 6, further comprising a history buffer configured to store one or more of the plurality of features.

8. The apparatus of claim 6, wherein the relevance parameter is generated by one or more of:

a correlation, a probability, and a distance measure between an occurrence of the plurality of features and an occurrence of the motor control signal.

9. The apparatus of claim 8, wherein:

the first characteristic is selected from a group consisting of;

a feature color, a feature texture, a feature edge orientation, and a feature type; and

the second characteristic comprises a feature motion.

10. The apparatus of claim 6, wherein:

the first transformation process comprises a sparsification operation configured to select one or more first features from a plurality of first outputs produced by the first transformation process based on the first characteristic; and

the sparsification operation is further configured to reduce a number of data elements by a factor of at least four.

11. The apparatus of claim 10, wherein the first characteristic comprises a statistical parameter further determined based on the plurality of first outputs of the first transformation process.

12. The apparatus of claim 11, wherein the statistical parameter comprises one or more of:

a mean, a variance, a percentile, and a standard deviation.

13. A system for determining a motor control signal for a robotic apparatus based on an analysis of an external input, the system comprising:

a sensor configured to receive the external input;

a non-transitory computer-readable memory configured to store a plurality of computer instructions; and

at least one processing component configured to cause the plurality of computer instructions to, when executed;

analyze the external input via a first transformation process applied based on a presence of a first set of features from the input signal with a first characteristic;

analyze the external input via a second transformation process applied based on a presence of a second set of features from the external input with a second characteristic;

determine a plurality of features based on the analysis of the external input via the first transformation process over a first period of time and the analysis of the external input via the second transformation process over a second period of time;

normalize each of the plurality of features using a statistical parameter determined from the plurality of features;

provide a plurality of normalized features to a classification process configured to detect one or more relevant features among the plurality of normalized features, the one or more relevant features comprising a relevance value that exceeds a threshold; and

select from a plurality of control signals a motor control signal associated with the one or more relevant features with the relevance value that exceeds the threshold;

wherein the selection of the motor control signal is based on a previous concurrent occurrence of the motor control signal and the one or more relevant features; and

wherein the robotic apparatus is configured to execute a task based on the selection of the motor control signal.

14. The system of claim 13, wherein:

the external input comprises a training control indication;

the at least one processing component is further configured to operate a combiner process configured to;

produce a combined control indication based at least on the training control indication; and

provide the combined control indication;

wherein the motor control signal is determined at least in part on the combined control indication.

15. The system of claim 14, wherein:

the combiner process comprises an override transfer function; and

a non-zero value of the training control indication is configured to cause a determination of the combined control indication.

16. The system of claim 13, wherein a relevance of the one or more relevant features comprises a parameter generated from one or more of:

a correlation, a probability, and a distance measure between the concurrent occurrence of the motor control signal and the one or more relevant features.

17. The system of claim 13, wherein:

the first characteristic is selected from a group consisting of;

a feature color, a feature texture, a feature edge orientation, and a feature type; and

the second characteristic characterizes a feature motion.

18. The system of claim 13, wherein:

the first transformation process comprises a sparsification operation comprising selecting one or more first features from a plurality of first outputs produced by the first transformation process based on the first characteristic; and

the sparsification operation is further configured to reduce a number of data elements by a factor of at least four.

19. The system of claim 13, wherein the first characteristic comprises a statistical parameter further determined based on the plurality of first outputs of the first transformation process.