APPARATUS AND METHODS FOR TRAINING OF ROBOTIC CONTROL ARBITRATION
First Claim
1. A processor-implemented method of learning arbitration for two tasks by a controller of a robot, the method being performed by one or more processors configured to execute computer program modules, the method comprising:
- during a first training trial of a plurality of trials;
evaluating, using one or more processors, a first performance associated with contemporaneous execution of the two tasks by the robot, the two tasks including a first task and a second task;
based on the first performance being within a range from a second performance obtained during a second training trial of the plurality of trials and prior to the first training trial, receiving a teaching signal conveying information associated with a target task, the target task being one of the two tasks; and
based on the teaching signal, determining an association between the context and the target task, the association being configured to cause the robot to execute the target task based on (1) an occurrence of the context after the first training trial during a third training trial of the plurality of trials, and (2) an absence of receiving the teaching signal during the third training trial.
1 Assignment
0 Petitions
Accused Products
Abstract
Apparatus and methods for arbitration of control signals for robotic devices. A robotic device may comprise an adaptive controller comprising a plurality of predictors configured to provide multiple predicted control signals based on one or more of the teaching input, sensory input, and/or performance. The predicted control signals may be configured to cause two or more actions that may be in conflict with one another and/or utilize a shared resource. An arbitrator may be employed to select one of the actions. The selection process may utilize a WTA, reinforcement, and/or supervisory mechanisms in order to inhibit one or more predicted signals. The arbitrator output may comprise target state information that may be provided to the predictor block. Prior to arbitration, the predicted control signals may be combined with inputs provided by an external control entity in order to reduce learning time.
109 Citations
20 Claims
-
1. A processor-implemented method of learning arbitration for two tasks by a controller of a robot, the method being performed by one or more processors configured to execute computer program modules, the method comprising:
during a first training trial of a plurality of trials; evaluating, using one or more processors, a first performance associated with contemporaneous execution of the two tasks by the robot, the two tasks including a first task and a second task; based on the first performance being within a range from a second performance obtained during a second training trial of the plurality of trials and prior to the first training trial, receiving a teaching signal conveying information associated with a target task, the target task being one of the two tasks; and based on the teaching signal, determining an association between the context and the target task, the association being configured to cause the robot to execute the target task based on (1) an occurrence of the context after the first training trial during a third training trial of the plurality of trials, and (2) an absence of receiving the teaching signal during the third training trial. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13)
-
14. A processor-implemented method of selecting a motor actuator signal from a first actuator signal and a second actuator signal, the method being performed by one or more processors configured to execute computer program modules, the method comprising:
-
based on a sensory context, providing the first actuator signal and the second actuator signal, the first actuator signal being configured to effectuate execution of a first action and the second actuator signal being configured to effectuate execution of a second action; at a first time instance, contemporaneously applying the first actuator signal and the second actuator signal to a motor; determining, using one or more processors, a first performance value associated with the contemporaneous application of the first actuator signal and the second actuator signal; and based on a comparison of the first performance value to a second performance value, selecting the first actuator signal or the second actuator signal as the motor actuator signal; wherein; the comparison is indicative of a mismatch between of the first performance value and the second performance value; and the second performance value is associated with another contemporaneous application of the first actuator signal and the second actuator signal to the motor at a second time instance prior to the first time instance. - View Dependent Claims (15, 16, 17, 18)
-
-
19. A non-transitory computer-readable storage medium having instructions embodied thereon, the instructions being executable by a processor to perform a method of predicting a control signal associated with a robotic apparatus, the method comprising:
-
detecting an object in sensory input provided by a sensor of the robotic apparatus; determining a control output based on a characteristic of the object; operating the robotic apparatus in accordance with the control output, the control output being capable of causing the robotic apparatus to execute a maneuver that accounts for the characteristic; determining a predicted control output based on the characteristic; determining a combined output based on the control output and the predicted control output, the combined output being characterized by a transform function; based on an error measure, withdrawing control output from the combined output to produce the predicted control output; and providing the predicted control output to the robotic apparatus, the predicted control output being capable of causing the execution of the maneuver. - View Dependent Claims (20)
-
Specification