Method and system for dynamic task selection suitable for mapping external inputs and internal goals toward actions that solve problems or elicit rewards

• US 8,762,305 B1
• Filed: 11/02/2011
• Issued: 06/24/2014
• Est. Priority Date: 11/11/2010
• Status: Active Grant

• Current Assignee: HRL Laboratories LLC (The Boeing Co.)
• Original Assignee: HRL Laboratories LLC (The Boeing Co.)
• Inventors: Srinivasa, Narayan, Chelian, Suhas E.
• Primary Examiner(s): Gaffin, Jeffrey A
• Assistant Examiner(s): BUSS, BENJAMIN J
• Application Number:
  US13/287,953
• Time in Patent Office: 965 Days
• Field of Search: 706/12, 706/14, 706/45, 706/48, 706/903
• US Class Current: 706/14
• CPC Class Codes: G06N 20/00 Machine learning
• 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 3/08 Learning methods
• CPC Class Codes: Y10S 706/903 Control

Abstract

The present invention relates to a system for mapping external inputs and internal goals toward actions that solve problems or elicit external rewards. The present invention allows an instructor to test and train an agent to perform dynamic task selection (executive control) by using a schema that computes the agent'"'"'s emotional and motivational states from reward/punishment inputs and sensory inputs (visual, auditory, kinematic, tactile, olfactory, somatosensory, and motor inputs). Specifically, the invention transforms the sensory inputs into unimodal and bimodal spatio-temporal schemas that are combined with the reward/punishment inputs and with the emotional and motivation states to create an external/internal schema (EXIN schema), that provides a compressed representation assessing the agent'"'"'s emotions, motivations, and rewards. The invention uses the EXIN schema to create a motor schema to be executed by the agent to dynamically perform the task selected by the instructor.

Claims

1. A system for dynamic task selection (executive control) suitable for mapping external inputs and internal goals toward actions that solve problems or elicit external rewards, the system comprising a processor, an input coupled with the processor for receiving a plurality of inputs to be processed into a motor-schema, and a memory coupled with the processor, wherein the memory includes instruction means that are executable by the processor for causing the processor to perform operations of:

receiving sensory inputs;

transforming the sensory inputs into clusters of spatial patterns;

transforming the clusters of spatial patterns into a spatial schema;

encoding transitions between the spatial schema;

clustering the spatial schema and encoded spatial schema transitions into a spatio-temporal schema;

combining the spatial schema with the spatio-temporal schema to create a bimodal spatio-temporal schema;

receiving a reward input and a punishment input, wherein the reward input and the punishment input reflect a current state of an external environment;

computing an emotional state and a motivational state from the sensory inputs, the reward input, and the punishment input;

combining the bimodal spatio-temporal schema with the reward input, the punishment input, the emotional state, and the motivational state to create an external/internal schema (EXIN schema), wherein the EXIN schema provides a compressed representation assessing emotions, motivations, and rewards;

combining the EXIN schema with the bimodal spatio-temporal schema to create a multimodal spatio-temporal schema, wherein the multimodal spatio-temporal schema serves as an episodic memory that can be replayed by the system;

receiving structures representing a plurality of elements of a motor system; and

combining the multimodal spatio-temporal schema with the motor system structures and the EXIN schema to create a motor schema.

2. A system for dynamic task selection as set forth in claim 1, wherein the system performs in two distinct modes, a training mode and a testing mode:

wherein in the training mode, the created motor schema represents a learnt motor schema associated with emotional and motivational-drive context derived from the EXIN schema following a bottom-up input hierarchy, and the processor further performs the operation of storing the learnt motor schema; and

wherein in the testing mode, the created motor schema indicates how to perform a particular action, and the processor further performs the operation of feeding back the EXIN schema to be combined with the spatial schema and the spatio-temporal schema to create a top-down bimodal spatio-temporal schema, and the processor further performs the operation of sending the created motor schema to the motor system to perform a particular action.

3. A system for dynamic task selection as set forth in claim 2, wherein the system further comprises an agent to be trained to perform dynamic task selection, an instructor for providing a task to the agent to perform, and a data base of motor schemas that fulfill a plurality of tasks to be selected by the instructor, and wherein the processor further performs the operations of:

receiving from the instructor a desired task to be performed by the agent;

selecting from the data base a desired motor schema that fulfills the desired task received from the instructor;

executing the motor schema created by the system;

comparing a difference measurement between the executed motor schema and the desired motor schema;

computing a progress measurement of the agent from the reward input and the schema difference measurement, wherein the progress measurement increases as the executed motor schema gets closer to the desired motor schema and closer to achieving the desired task, and wherein the progress measurement decreases as the executed motor schema gets farther away from the desired motor schema;

generating a reward signal when the progress measurement increases, indicating that the agent is closer to achieve the desired task;

keeping a count of number of reward signals generated by the system to compute reward frequency associated with the created motor schema;

generating a punishment signal when the progress measurement decreases, indicating that the agent is not achieving the desired task;

keeping a count of number of punishment signals generated by the system; and

feeding back the reward signal and punishment signal to be inputted into the system as reward input and punishment input, respectively.

4. A system for dynamic task selection as set forth in claim 3, wherein the processor further performs the operations of:

receiving a mismatch threshold from the instructor;

comparing a mismatch threshold with the count of punishment signals and generating a mismatch reset signal when the count of punishment signals exceeds the mismatch threshold; and

sending the mismatch reset signal to the multimodal pre-processor, wherein mismatch reset signal forces the multimodal pre-processor to deselect the EXIN schema and to further perform the operation of creating a new EXIN schema; and

sending the mismatch reset signal to the motor system to shut-off motor channels and stop the motor system from performing the particular action being performed.

5. A system for dynamic task selection as set forth in claim 4, wherein the processor further performs the operations of:

combining the sensory inputs with the count of reward signals to create a reset toggle signal when the count of reward signals is drastically reduced or the sensory inputs reflect significantly changed emotional states; and

sending the reset toggle signal to the multimodal pre-processor, wherein reset toggle signal forces the multimodal pre-processor to stop creating the EXIN schema; and

sending the reset toggle signal to the motor system to shut-off motor channels and stop the motor system from performing actions.

6. A system for dynamic task selection as set forth in claim 5, wherein the processor further performs the operations of:

receiving a success threshold from the instructor;

using the count of reward signals to compute the reward frequency associated with the executed motor schema; and

comparing the success threshold with reward frequency of the executed motor schema and storing the executed motor schema as a habit or skill when the reward frequency meets the success threshold required by the instructor.

7. A system for dynamic task selection as set forth in claim 6, wherein the reward input is further used as an enticing reset signal when the punishment input has a high value, whereby the enticing reset signal forces the system to deselect the EXIN schemas and continue creating new EXIN schemas despite the high value of the punishment input.

8. A system for dynamic task selection as set forth in claim 7, wherein the processor further performs the operations of:

calculating set of statistics reflecting primacy/recency, frequency of occurrence, and reward and punishment associated with the biomodal spatio-temporal schema, with the multimodal spatio-temporal schema, and with the motor schema;

analyzing the set of statistics to resolve conflicts generated by a suddenly changing desired task or suddenly changing sensory inputs;

creating a conflict resolution signal based on the set of statistics, wherein the conflict resolution signal selects, to be processed, schemas that are more frequently executed and that have high reward signals associated with the schemas, and also selects to be processed schemas that are more recently executed than older schemas.

9. A system for dynamic task selection as set forth in claim 8, wherein the EXIN schema is created by further combining the spatio-temporal schema with the bimodal spatio-temporal schema, the reward input, the punishment input, the emotional state, and the motivational state;

wherein the motor schema is created by further combining the spatial schema and the bimodal spatio-temporal schema with the multimodal spatio-temporal schema, the motor system structures, and the EXIN schema; and

wherein the sensory inputs are selected from a group consisting of;

visual inputs, auditory inputs, kinematic inputs, tactile inputs, olfactory inputs, somatosensory inputs, and motor inputs.

10. A system for dynamic task selection as set forth in claim 9, wherein the operations performed by the processor are defined by differential equations with timing constraints that are implemented on a serial computer or on a parallel Analog Pulse Processing computer, wherein the EXIN schema corresponds to compressed representations of an EXternal motivator and INternal states and the motor schema corresponds to a sequence of joint rotations, where space corresponds to the joint rotations and time corresponds to a course taken by the joint rotations along the particular action;

and wherein the EXIN schemas, the motor schemas, and the spatio-temporal schemas are encoded and recognized via a spatio-temporal algorithm selected from a group consisting of;

hierarchical Adaptive Resonance Theory Store (hARTSTORE) network, hierarchical temporal memory (HTM), sequence of clusters (SOC) and deep belief network (DBN).

11. A system for dynamic task selection as set forth in claim 1, wherein the system further comprises an agent to be trained to perform dynamic task selection, an instructor for providing a task to the agent to perform, and a data base of motor schemas that fulfill a plurality of tasks to be selected by the instructor, and wherein the processor further performs the operations of:

receiving from the instructor a desired task to be performed by the agent;

selecting from the data base a desired motor schema that fulfills the desired task received from the instructor;

executing the motor schema created by the system;

comparing a difference measurement between the executed motor schema and the desired motor schema;

computing a progress measurement of the agent from the reward input and the schema difference measurement, wherein the progress measurement increases as the executed motor schema gets closer to the desired motor schema and closer to achieving the desired task, and wherein the progress measurement decreases as the executed motor schema gets farther away from the desired motor schema;

generating a reward signal when the progress measurement increases, indicating that the agent is closer to achieve the desired task;

keeping a count of number of reward signals generated by the system to compute reward frequency associated with the created motor schema;

generating a punishment signal when the progress measurement decreases, indicating that the agent is not achieving the desired task;

keeping a count of number of punishment signals generated by the system; and

feeding back the reward signal and punishment signal to be inputted into the system as reward input and punishment input, respectively.

12. A system for dynamic task selection as set forth in claim 11, wherein the processor further performs the operations of:

receiving a mismatch threshold from the instructor;

comparing mismatch threshold with the count of punishment signals and generating a mismatch reset signal when the count of punishment signals exceeds the mismatch threshold; and

sending the mismatch reset signal to the multimodal pre-processor, wherein mismatch reset signal forces the multimodal pre-processor to deselect the EXIN schema and to further perform the operation of creating a new EXIN schema; and

sending the mismatch reset signal to the motor system to shut-off motor channels and stop the motor system from performing the particular action being performed.

13. A system for dynamic task selection as set forth in claim 11, wherein the processor further performs the operations of:

combining the sensory inputs with the count of reward signals to create a reset toggle signal when the count of reward signals is drastically reduced or the sensory inputs reflect significantly changed emotional states; and

sending the reset toggle signal to the multimodal pre-processor, wherein reset toggle signal forces the multimodal pre-processor to stop creating the EXIN schema; and

sending the reset toggle signal to the motor system to shut-off motor channels and stop the motor system from performing actions.

14. A system for dynamic task selection as set forth in claim 11, wherein the processor further performs the operations of:

receiving a success threshold from the instructor,

using the count of reward signals to compute the reward frequency associated with the executed motor schema; and

comparing the success threshold with reward frequency of the executed motor schema and storing the executed motor schema as a habit or skill when the reward frequency meets the success threshold required by the instructor.

15. A system for dynamic task selection as set forth in claim 1, wherein the reward input is further used as an enticing reset signal when the punishment input has a high value, whereby the enticing reset signal forces the system to deselect the EXIN schemas and continue creating new EXIN schemas despite the high value of the punishment input.

16. A system for dynamic task selection as set forth in claim 1, wherein the processor further performs the operations of:

calculating set of statistics reflecting primacy/recency, frequency of occurrence, and reward and punishment associated with the biomodal spatio-temporal schema, with the multimodal spatio-temporal schema, and with the motor schema;

analyzing the set of statistics to resolve conflicts generated by a suddenly changing desired task or suddenly changing sensory inputs;

creating a conflict resolution signal based on the set of statistics, wherein the conflict resolution signal selects, to be processed, schemas that are more frequently executed and that have high reward signals associated with the schemas, and also selects to be processed schemas that are more recently executed than older schemas.

17. A system for dynamic task selection as set forth in claim 1, wherein the EXIN schema is created by further combining the spatio-temporal schema with the bimodal spatio-temporal schema, the reward input, the punishment input, the emotional state, and the motivational state;

wherein the motor schema is created by further combining the spatial schema and the bimodal spatio-temporal schema with the multimodal spatio-temporal schema, the motor system structures, and the EXIN schema; and

wherein the sensory inputs are selected from a group consisting of;

visual inputs, auditory inputs, kinematic inputs, tactile inputs, olfactory inputs, somatosensory inputs, and motor inputs.

18. A system for dynamic task selection as set forth in claim 1, wherein the operations performed by the processor are defined by differential equations with timing constraints that are implemented on a serial computer or on a parallel Analog Pulse Processing computer;

wherein the EXIN schema corresponds to compressed representations of an EXternal motivator and INternal states and the motor schema corresponds to a sequence of joint rotations, where space corresponds to the joint rotations and time corresponds to a course taken by the joint rotations along the particular action; and

wherein the EXIN schemas, the motor schemas, and the spatio-temporal schemas are encoded and recognized via a spatio-temporal algorithm selected from a group consisting of;

hierarchical Adaptive Resonance Theory Store (hARTSTORE) network, hierarchical temporal memory (HTM), sequence of clusters (SOC), and deep belief network (DBN).

19. A computer implemented method for dynamic task selection (executive control) suitable for mapping external inputs and internal goals toward actions that solve problems or elicit external rewards, the method comprising an act of causing a processor to execute instructions encoded on a memory, such that upon execution, the processor perform operations of:

receiving sensory inputs;

transforming the sensory inputs into clusters of spatial patterns;

transforming the clusters of spatial patterns into a spatial schema;

encoding transitions between the spatial schema;

clustering the spatial schema and encoded spatial schema transitions into a spatio-temporal schema;

combining the spatial schema with the spatio-temporal schema to create a bimodal spatio-temporal schema;

receiving a reward input and a punishment input, wherein the reward input and the punishment input reflect a current state of an external environment;

computing an emotional state and a motivational state from the sensory inputs, the reward input, and the punishment input;

combining the bimodal spatio-temporal schema with the reward input, the punishment input, the emotional state, and the motivational state to create an external/internal schema (EXIN schema), wherein the EXIN schema provides a compressed representation assessing emotions, motivations, and rewards;

combining the EXIN schema with the bimodal spatio-temporal schema to create a multimodal spatio-temporal schema, wherein the multimodal spatio-temporal schema serves as an episodic memory that can be replayed by the computer implemented method;

receiving structures representing a plurality of elements of a motor system; and

combining the multimodal spatio-temporal schema with the motor system structures and the EXIN schema to create a motor schema.

20. A computer implemented method for dynamic task selection as set forth in claim 19, wherein the method behaves in two distinct modes, a training mode and a testing mode:

wherein in the training mode, the created motor schema represents a learnt motor schema associated with emotional and motivational-drive context derived from the EXIN schema following a bottom-up input hierarchy, and the processor further performs the operation of storing the learnt motor schema; and

wherein in the testing mode, the created motor schema indicates how to perform a particular action, and the processor further performs the operation of feeding back the EXIN schema to be combined with the spatial schema and the spatio-temporal schema to create a top-down bimodal spatio-temporal schema, and the processor further performs the operation of sending the created motor schema to the motor system to perform a particular action.

21. A computer implemented method for dynamic task selection as set forth in claim 20, wherein the method is used by an agent to be trained to perform dynamic task selection along with an instructor that provides a task to the agent to perform, and wherein the method uses a data base of motor schemas that fulfill a plurality of tasks to be selected by the instructor, and wherein the processor further performs the operations of:

receiving from the instructor a desired task to be performed by the agent;

selecting from the data base a desired motor schema that fulfills the desired task received from the instructor;

executing the motor schema previously created by the operations performed by the processor;

comparing a difference measurement between the executed motor schema and the desired motor schema;

computing a progress measurement of the agent from the reward input and the schema difference measurement, wherein the progress measurement increases as the executed motor schema gets closer to the desired motor schema and closer to achieving the desired task, and wherein the progress measurement decreases as the executed motor schema gets farther away from the desired motor schema;

generating a reward signal when the progress measurement increases, indicating that the agent is closer to achieve the desired task;

keeping a count of number of reward signals generated by the computer implemented method to compute reward frequency associated with the created motor schema;

generating a punishment signal when the progress measurement decreases, indicating that the agent is not achieving the desired task;

keeping a count of number of punishment signals generated by the computer implemented method; and

feeding back the reward signal and punishment signal to be inputted into the computer implemented method as reward input and punishment input, respectively.

22. A computer implemented method for dynamic task selection as set forth in claim 21, wherein the processor further performs the operations of:

receiving a mismatch threshold from the instructor;

comparing mismatch threshold with the count of punishment signals and generating a mismatch reset signal when the count of punishment signals exceeds the mismatch threshold; and

sending the mismatch reset signal to the multimodal pre-processor, wherein mismatch reset signal forces the multimodal pre-processor to deselect the EXIN schema and to further perform the operation of creating a new EXIN schema; and

sending the mismatch reset signal to the motor system to shut-off motor channels and stop the motor system from performing the particular action being performed.

23. A computer implemented method for dynamic task selection as set forth in claim 22, wherein the processor further performs the operations of:

combining the sensory inputs with the count of reward signals to create a reset toggle signal when the count of reward signals is drastically reduced or the sensory inputs reflect significantly changed emotional states; and

sending the reset toggle signal to the multimodal pre-processor, wherein reset toggle signal forces the multimodal pre-processor to stop creating the EXIN schema; and

sending the reset toggle signal to the motor system to shut-off motor channels and stop the motor system from performing actions.

24. A computer implemented method for dynamic task selection as set forth in claim 23, wherein the processor further performs the operations of:

receiving a success threshold from the instructor;

using the count of reward signals to compute the reward frequency associated with the executed motor schema; and

comparing the success threshold with reward frequency of the executed motor schema and storing the executed motor schema as a habit or skill when the reward frequency meets the success threshold required by the instructor.

25. A computer implemented method for dynamic task selection as set forth in claim 24, wherein the reward input is further used as an enticing reset signal when the punishment input has a high value, whereby the enticing reset signal forces the computer implemented method to deselect the EXIN schemas and continue creating new EXIN schemas despite the high value of the punishment input.

26. A computer implemented method for dynamic task selection as set forth in claim 25, wherein the processor further performs the operations of:

calculating set of statistics reflecting primacy/recency, frequency of occurrence, and reward and punishment associated with the biomodal spatio-temporal schema, with the multimodal spatio-temporal schema, and with the motor schema;

analyzing the set of statistics to resolve conflicts generated by a suddenly changing desired task or suddenly changing sensory inputs;

creating a conflict resolution signal based on the set of statistics, wherein the conflict resolution signal selects, to be processed, schemas that are more frequently executed and that have high reward signals associated with the schemas, and also selects to be processed schemas that are more recently executed than older schemas.

27. A computer implemented method for dynamic task selection as set forth in claim 26, wherein the EXIN schema is created by further combining the spatio-temporal schema with the bimodal spatio-temporal schema, the reward input, the punishment input, the emotional state, and the motivational state;

wherein the motor schema is created by further combining the spatial schema and the bimodal spatio-temporal schema with the multimodal spatio-temporal schema, the motor system structures, and the EXIN schema; and

wherein the sensory inputs are selected from a group consisting of;

visual inputs, auditory inputs, kinematic inputs, tactile inputs, olfactory inputs, somatosensory inputs, and motor inputs.

28. A computer implemented method for dynamic task selection as set forth in claim 27, wherein the operations performed by the processor are defined by differential equations with timing constraints that are implemented on a serial computer or on a parallel Analog Pulse Processing computer;

wherein the EXIN schema corresponds to compressed representations of an EXternal motivator and INternal states and the motor schema corresponds to a sequence of joint rotations, where space corresponds to the joint rotations and time corresponds to a course taken by the joint rotations along the particular action; and

wherein the EXIN schemas, the motor schemas, and the spatio-temporal schemas are encoded and recognized via a spatio-temporal algorithm selected from a group consisting of;

hierarchical Adaptive Resonance Theory Store (hARTSTORE) network, hierarchical temporal memory (HTM), sequence of clusters (SOC), and deep belief network (DBN).

29. A computer implemented method for dynamic task selection as set forth in claim 19, wherein the computer implemented method is used by an agent to be trained to perform dynamic task selection and by an instructor that provides a task to the agent to perform, and wherein the method uses a data base of motor schemas that fulfill a plurality of tasks to be selected by the instructor, and wherein the processor further performs the operations of:

receiving from the instructor a desired task to be performed by the agent;

selecting from the data base a desired motor schema that fulfills the desired task received from the instructor;

executing the motor schema created by the computer implemented method;

comparing a difference measurement between the executed motor schema and the desired motor schema;

computing a progress measurement of the agent from the reward input and the schema difference measurement, wherein the progress measurement increases as the executed motor schema gets closer to the desired motor schema and closer to achieving the desired task, and wherein the progress measurement decreases as the executed motor schema gets farther away from the desired motor schema;

generating a reward signal when the progress measurement increases, indicating that the agent is closer to achieve the desired task;

keeping a count of number of reward signals generated by the computer implemented method to compute reward frequency associated with the created motor schema;

generating a punishment signal when the progress measurement decreases, indicating that the agent is not achieving the desired task;

keeping a count of number of punishment signals generated by the computer implemented method; and

feeding back the reward signal and punishment signal to be inputted into the computer implemented method as reward input and punishment input, respectively.

30. A computer implemented method for dynamic task selection as set forth in claim 29, wherein the processor further performs the operations of:

receiving a mismatch threshold from the instructor;

comparing mismatch threshold with the count of punishment signals and generating a mismatch reset signal when the count of punishment signals exceeds the mismatch threshold; and

sending the mismatch reset signal to the multimodal pre-processor, wherein mismatch reset signal forces the multimodal pre-processor to deselect the EXIN schema and to further perform the operation of creating a new EXIN schema; and

sending the mismatch reset signal to the motor system to shut-off motor channels and stop the motor system from performing the particular action being performed.

31. A computer implemented method for dynamic task selection as set forth in claim 29, wherein the processor further performs the operations of:

combining the sensory inputs with the count of reward signals to create a reset toggle signal when the count of reward signals is drastically reduced or the sensory inputs reflect significantly changed emotional states; and

sending the reset toggle signal to the multimodal pre-processor, wherein reset toggle signal forces the multimodal pre-processor to stop creating the EXIN schema; and

sending the reset toggle signal to the motor system to shut-off motor channels and stop the motor system from performing actions.

32. A computer implemented method for dynamic task selection as set forth in claim 29, wherein the processor further performs the operations of:

receiving a success threshold from the instructor;

using the count of reward signals to compute the reward frequency associated with the executed motor schema; and

comparing the success threshold with reward frequency of the executed motor schema and storing the executed motor schema as a habit or skill when the reward frequency meets the success threshold required by the instructor.

33. A computer implemented method for dynamic task selection as set forth in claim 19, wherein the reward input is further used as an enticing reset signal when the punishment input has a high value, whereby the enticing reset signal forces the computer implemented method to deselect the EXIN schemas and continue creating new EXIN schemas despite the high value of the punishment input.

34. A computer implemented method for dynamic task selection as set forth in claim 19, wherein the processor further performs the operations of:

calculating set of statistics reflecting primacy/recency, frequency of occurrence, and reward and punishment associated with the biomodal spatio-temporal schema, with the multimodal spatio-temporal schema, and with the motor schema;

analyzing the set of statistics to resolve conflicts generated by a suddenly changing desired task or suddenly changing sensory inputs;

creating a conflict resolution signal based on the set of statistics, wherein the conflict resolution signal selects, to be processed, schemas that are more frequently executed and that have high reward signals associated with the schemas, and also selects to be processed schemas that are more recently executed than older schemas.

35. A computer implemented method for dynamic task selection as set forth in claim 19, wherein the EXIN schema is created by further combining the spatio-temporal schema with the bimodal spatio-temporal schema, the reward input, the punishment input, the emotional state, and the motivational state;

wherein the motor schema is created by further combining the spatial schema and the bimodal spatio-temporal schema with the multimodal spatio-temporal schema, the motor system structures, and the EXIN schema; and

wherein the sensory inputs are selected from a group consisting of;

visual inputs, auditory inputs, kinematic inputs, tactile inputs, olfactory inputs, somatosensory inputs, and motor inputs.

36. A computer implemented method for dynamic task selection as set forth in claim 19, wherein the operations performed by the processor are defined by differential equations with timing constraints that are implemented on a serial computer or on a parallel Analog Pulse Processing computer;

wherein the EXIN schema corresponds to compressed representations of an EXternal motivator and INternal states and the motor schema corresponds to a sequence of joint rotations, where space corresponds to the joint rotations and time corresponds to a course taken by the joint rotations along the particular action; and

wherein the EXIN schemas, the motor schemas, and the spatio-temporal schemas are encoded and recognized via a spatio-temporal algorithm selected from a group consisting of;

hierarchical Adaptive Resonance Theory Store (hARTSTORE) network, hierarchical temporal memory (HTM), sequence of clusters (SOC), and deep belief network (DBN).