FEATURE DETECTION APPARATUS AND METHODS FOR TRAINING OF ROBOTIC NAVIGATION

• US 20160096270A1
• Filed: 11/14/2014
• Published: 04/07/2016
• Est. Priority Date: 10/02/2014
• Status: Active Grant

• Current Assignee: Brain Corporation
• Original Assignee: Brain Corporation
• Inventors: Smith, Andrew, O'Connor, Peter, Ibarz Gabardos, Borja
• Granted Patent: US 9,630,318 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• 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 operating a robotic device by a learning controller, the method comprising:

during an operation of a plurality of operations;

causing the robotic device to execute an action along a first trajectory in accordance with a first control signal determined based on a sensory input;

determining, a performance measure based on an evaluation of the first trajectory and an indication related to a target trajectory provided by a trainer;

conveying information related to the performance measure to the learning controller; and

updating one or more learning parameters of the learning controller in accordance with the information; and

during a subsequent operation of the plurality of operations;

causing the robotic device to execute the action along a second trajectory in accordance with a second control signal determined based on the sensory input;

wherein;

the execution of the action along the second trajectory is characterized by a second performance measure; and

the updating is configured to displace the second trajectory closer towards the target trajectory relative to the first trajectory.

2. The method of claim 1, wherein:

the first control signal is based on a plurality of features being detected in the sensory input by a feature detection component comprising a first plurality of spiking neuron nodes; and

the updating is configured to modify one or more parameters associated with one or more first nodes of the first plurality of spiking neuron nodes.

3. The method of claim 2, wherein:

the first control signal is determined by one or more second nodes of a second plurality of spiking neuron nodes;

the one or more first nodes of the first plurality of nodes are coupled to the one or more second nodes of the second plurality of nodes via a connectivity array of efficacies; and

the updating comprises a plasticity operation configured to modify one or more efficacies of the connectivity array of efficacies.

4. The method of claim 3, wherein:

individual first ones of the first plurality of nodes are coupled to individual second ones of the second plurality of nodes via an all-to-all connectivity pattern; and

the connectivity array of efficacies is characterized by a first dimension determined based on a first number of the one or more first nodes of the first plurality of nodes and a second dimension determined based on a second number of the one or more second nodes of the second plurality of nodes.

5. The method of claim 1, wherein:

the performance measure comprises a first distance between the first trajectory and the target trajectory; and

the second performance measure comprises a second distance between the second trajectory and the target trajectory, the second distance being smaller than the first distance.

6. The method of claim 1, wherein:

the performance measure comprises a first probability parameter between the first trajectory and the target trajectory; and

the second performance measure comprises a second probability parameter between the second trajectory and the target trajectory, the second probability parameter being greater than the first probability.

7. The method of claim 1, wherein:

the learning controller is configured for operation in accordance with a supervised learning process configured based on a teaching signal; and

the first control signal comprises a combination of the second layer output and a teaching signal provided to the robotic device.

8. The method of claim 1, wherein:

the operation and the another operation occur at a part of an online learning process wherein within a given time period, the robot may receive a teaching input, modify the learning process based on the teaching input and the learning process configuration, and subsequently navigate the trajectory based on the modified learning process.

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

analyzing the sensor input using a first transformation process configured by one or more first features present in the sensor input using a first characteristic;

analyzing the sensor input using a second transformation process configured by one or more second features present in the sensor input using a second characteristic, the second characteristic being different from the first characteristic;

determining a feature history comprising a plurality of features determined based on analysis of the sensor input over a time interval;

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

providing a plurality of normalized features to a classification process configured to detect at least one relevant feature among the plurality of normalized features, the relevant feature being characterized by a greater relevancy parameter compared to other ones of the plurality of normalized features; and

selecting from a plurality of control signals the motor control signal associated with the relevant feature;

wherein the selection of the motor control signal is effectuated based on prior contemporaneous occurrence of the motor control signal and the relevant feature.

10. The method of claim 9, wherein the relevance parameter comprises one or more of a correlation, a probability, and a distance measure between an occurrence of the individual ones of the plurality of features and an occurrence of the motor signal.

11. The method of claim 10, wherein:

the first characteristic is selected from a group consisting of feature color, feature texture, feature edge orientation, and feature type; and

the second characteristic is configured to characterize a feature motion.

12. The method of claim 9, 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 based on the first characteristic; and

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

13. The method of claim 12, wherein the first characteristic comprises a statistical parameter determined based on a plurality of outputs of the transformation.

14. A self-contained robotic apparatus the apparatus comprising:

a platform comprising a motor;

a sensor component configured to obtain a plurality of video frames at a frame rate;

one or more processors configured to operate;

a feature detection logic configured to detect a plurality of features based on analysis of the plurality of video frames; and

a learning controller logic configured to produce a control indication based on the plurality of features and a performance indication;

an adapter component configured to produce a motor activation signal based on the control indication; and

an enclosure configured to house the motor, the sensor component, the adaptor component, and the one or more processors;

wherein;

the motor activation signal is configured to cause the self-contained robotic apparatus to execute an action characterized by an outcome;

the one or more processors is characterized by a processing capacity being sufficient for operate the feature detection and the learning component logic so that for a given frame of the plurality of frames, the control indication is being produced based on one or more features corresponding to the given frame prior to occurrence of a subsequent frame of the plurality of video frames;

the feature detection is configured based on application of a plurality of spatial filter operations to individual frames of the plurality of video frames; and

the performance indication is configured based on the outcome and a target outcome.

15. The apparatus of claim 14, further comprising a communications interface configured to receive an external input indicative of the target outcome.

16. The apparatus of claim 15, wherein:

the control indication for the given frame of the plurality of frames is configured based on another previously determined control indication produced based on a preceding frame relative to the given frame and the external input.

17. The apparatus of claim 16, wherein:

the external input comprises a training control indication;

the one or more processors is further configured to operate a combiner process configured to produce a combined control indication based on the control indication and the training control indication; and

provision of the combined control indication is configured to cause determination of the motor activation signal.

18. The apparatus of claim 17, wherein:

the combiner process is characterized by an override transfer function wherein a non-zero value of the training control indication is configured to cause determination of the combined control indication independent from the control indication.

19. The apparatus of claim 17, wherein:

the combiner process is characterized by an additive transfer function wherein the combined control indication configured based on a combination of the control indication the training control indication.

20. The apparatus of claim 17, wherein:

the combiner process is characterized by an listen transfer function wherein the combined control indication configured based solely on the control indication; and

the learning process adaptation is configured based on the training control indication.

21. The apparatus of claim 16, wherein:

the learning process comprises a nearest neighbor classifier process configured to;

based on one or more features corresponding to one or more current input frames, determine a plurality of neighbors, individual ones of the plurality of neighbors comprising (i) one or more prior features determined based on one or more input frames analyzed at a prior time previously to the current input frame; and

(ii) a prior control indication produced based on an evaluation of the one or more prior features at the prior time; and

the control indication is configured based on a combination of individual ones of a plurality of prior control indications associated with individual ones of the plurality of neighbors.